Intelligent Cyber Physical Systems and Internet of Things: ICoICI 2022 3031184963, 9783031184963

Table of contents :
Preface
Contents
Term Frequency Tokenization for Fake News Detection
1 Introduction
2 Literature Review
3 Methodology
4 Results and Discussion
5 Conclusion
References
Aquaculture Monitoring System Using Internet of Things
1 Introduction
2 Related Works
3 Methodology
3.1 Data Extraction
3.2 Data Telemetry and Visualization
3.3 Parameter Prediction
3.4 Chatbot
3.5 Water Quality Maintenance
4 Design and Implementation
4.1 Hardware Details
4.2 Software Details
4.3 Overview of Implementation
4.4 Implementation of pH Prediction
5 Results and Discussion
5.1 Sensor Data Visualization in IBM Watson Internet of Things Platform
5.2 Parameter Prediction
5.3 Chatbot
5.4 Android App
6 Conclusion and Future Scope
References
A Comprehensive Study and Implementation of Memory Malware Analysis with Its Application for the Case Study of CRIDEX
1 Introduction
2 Background Work
3 Related Work
4 Memory Malware Analysis Approach
5 Available Tools
6 CRIDEX Case Study
7 Conclusion
References
IoT Based Anti Poaching of Trees and Protection of Forest
1 Introduction
2 Related Works
3 Proposed Work
4 Results
5 Conclusion and Future Works
5.1 Conclusion
5.2 Future Works
References
Artificial Intelligence Based Efficient Activity Recognition with Real Time Implementation for ATM Security
1 Introduction
2 Literature Review
3 Proposed Model
4 Results and Dıscussıon
5 Conclusion
References
Terror Attack Classification with the Application of Orange Data Mining Tool and Neo4j Sandbox
1 Introduction
1.1 Dataset
1.2 Graph and Neo4j
2 Literature Survey
3 Methodology
3.1 Graph Creation
3.2 Graph Embedding Calculation
4 Model Building
4.1 Counting Null Values
4.2 Removal of Redundant Data
4.3 Dimension Reduction
4.4 Data Scaling
5 Results and Analysis
6 Conclusion
References
Multipurpose IoT Based Camera Using Deep Learning
1 Introduction
2 Related Work
3 Proposed Work
3.1 Hardware Integration
3.2 Training the Model
3.3 Detect and Predict Mask
3.4 Eye on Mask
3.5 Eye Open or Closed
3.6 Eye and Eye Ball Contours
3.7 Head Pose of Person
3.8 Function Calling and Display of Output
4 Result Analysis
5 Conclusion
References
Dr. Watson AI Based Healthcare Technology Project
1 Introduction
2 Methodology
3 Implementation
4 Results
4.1 Home Page
4.2 Dashboard Page
4.3 Medicine Prescriber
4.4 Diabetic Analyzer
4.5 Covid Predictor
4.6 Watson Assistant Chatbot
5 Conclusion
References
Empirical and Statistical Comparison of RSA and El-Gamal in Terms of Time Complexity
1 Introduction
2 Literature Review
3 Run Time Comparison of RSA and EL Gamal Protocols Used in MANET
4 Fundamental Theorem of Finite Difference
5 Result and Analysis
6 Conclusions and Future Work
References
IoT Communication to Capture and Store Data to Thingspeak Cloud Using NodeMCU and Ultrasonic Sensor
1 Introduction
2 Related Work
3 Proposed Work
4 Results and Discussion
5 Conclusion and Future Scope
References
A Comprehensive Study on Cloud Computing: Architecture, Load Balancing, Task Scheduling and Meta-Heuristic Optimization
1 Introduction
2 Cloud Computing: Architecture
2.1 Overview
2.2 Features of Cloud Computing
2.3 Architecture
2.4 Data Center in Cloud Network
2.5 Virtual Machines in Cloud Computing
2.6 Motivation of Study
3 Load Balancing in Cloud Computing
3.1 Classification of LB Techniques
3.2 Load Balancing Metrics
4 Task Scheduling in Cloud Environment
4.1 Types of Task Scheduling
4.2 Framework of TS in CC Environment
5 Meta-Heuristic Optimization Techniques
6 Literature Survey
6.1 Research Gaps and Issues
7 Conclusion
References
Balancing Exploration and Exploitation in Nature Inspired Computing Algorithm
1 Introduction
2 Existing Work
2.1 Particle Swarm Optimization
3 Proposed Work
3.1 Improved Multiple Particle Swarm Optimization (IMPSO)
4 Results and Discusssion
5 Conclusion
References
Blockchain Based Secure, Efficient, and Scalable Platform for the Organ Donation Process of Healthcare Industry
1 Introduction
2 Related Work
3 Preliminaries
3.1 Blockchain
3.2 Ethereum
3.3 Smart Contracts
4 The Proposed Organ Donation Platform Using Blockchain Technology
5 Results and Analysis
6 Conclusions and Future Works
References
Image Enhancement in Frequency Domain Fingerprint Detection and Matching Approach
1 Introduction
2 Frequency Domain Approach in Enhancement
3 Adopted Methodology
4 Design and Implementation
5 Simulation Results and Discussion
6 Conclusion
References
Developing Machine Learning Based Mobile App for Agriculture Application
1 Introduction
2 Relevance
3 Related Work
4 Proposed Work
5 Block Diagram
6 Architecture Description
7 Live Field Status
8 Crop Detection and Market Prediction
8.1 Data Gathering
8.2 Data Pre-processing
9 Crop Detection
9.1 Image Acquisition
9.2 Image Pre-processing
9.3 Feature Extraction
9.4 SVM Classification
9.5 Segmentation
10 Implementation and Result Analysis
10.1 Home Screen Along with Application Menu
10.2 Confusion Matrix
10.3 ML Model Comparison
10.4 Notification and Information
11 Conclusion
References
Attack Detection in IoT Using Machine Learning—A Survey
1 Introduction
2 IoT Architectures Attacks and Attacks Detection
3 Challenges
4 Conclusion
References
An Extensive Study on Logic Emerging IoT Adiabatic Techniques for Low-Power Circuit
1 Introduction
2 Background on Adiabatic Circuits
3 Recent Developments
3.1 Adiabatic Circuits for RF-Powered IoT Applications
3.2 Adiabatic Logic Enhanced Hardware Security
4 Avoid Common Mistakes to Adiabatic Logic Design
4.1 Don’t Use Diodes
4.2 Rules of Don’t Disobey Transistor
4.3 Use Reversible Logic
4.4 Don’t Over-Constrain the Design
5 Conclusion
References
A Critical Review of Agri-Food Supply Management with Traceability and Transparency Using Blockchain Technology
1 Introduction
1.1 Motivation and Contribution of Research Work
2 Related Work
2.1 Traceability
2.2 Efficiency
2.3 Privacy
2.4 Cost
2.5 Inventory Management
3 Blockchain Implementation Challenges
4 Conclusion
References
Face-Anti-spoofing Based on Liveness Detection
1 Introduction
2 Related Work
3 Proposed Methodology
3.1 Dataset Gathering
3.2 Data Pre-processing
3.3 Model Parameter
4 Simulation Result
4.1 Performance Matrix
4.2 Confusion Matrix
5 Conclusion
References
PDR Analysis and Network Optimization of Routing Protocols for Edge Networks
1 Introduction
1.1 Dynamic Source Routing (DSR)
1.2 Ad-hoc On-demand Distance Vector (AODV)
2 Related Work
3 Proposed Work
4 Result Analysis
5 Conclusion
References
Privacy Threat Reduction Using Modified Multi-line Code Generation Algorithm (MMLCGA) for Cancelable Biometric Technique (CBT)
1 Introduction
2 Related Work
3 Proposed Method
3.1 System Modules
3.2 System Architecture
4 Implementation Details
4.1 Binarization
4.2 Contextual Filtering
4.3 Thinning
4.4 Minutiae Extraction
4.5 Line Code Algorithm
5 Results and Discussion
6 Conclusion
References
Systematic Literature Review—IoT-Based Supply Chain Management in Industry 4.0
1 Introduction
2 IoT Based Supply Chain Management in Industry 4.0
3 Principles and Methods
3.1 Review Principles
3.2 Systematic Literature Review Method
4 Review of Research Trends
5 Content Analysis
6 Industry 4.0 and IoT in Supply Chain Research (Future Work)
7 Conclusion
References
A Review on Urban Flood Management Techniques for the Smart City and Future Research
1 Introduction
2 Related Work
3 Analysis and Discussion
4 Conclusion
References
Application of Distributed Constraint Optimization Technique for Privacy Preservation in Cyber-Physical Systems
1 Introduction
2 Related Work
3 Proposed Mechanism
4 Experimental Results
5 Future Work
6 Conclusion
References
Grip Assisting Glove for Charcot-Marie-Tooth Patients
1 Introduction
2 Literature Survey
2.1 Smart Muscle Strength Assessment Glove for Rehabilitation Purposes [7]
2.2 Calibration and Evaluation of a Force Measurement Glove for Field-Based Monitoring of Manual Wheelchair Users [8]
2.3 A Fabricated Force Glove that Measures Hand Forces During Activities of Daily Living [9]
2.4 Wearable System with Embedded Force Sensors for Neurological Rehabilitation Trainings [10]
2.5 Tactile Sensorized Glove for Force and Motion Sensing [11]
2.6 Smart Glove to Measure a Grip Force of the Workers [12]
2.7 Force-Sensing Glove System for Measurement of Hand Forces During Motorbike Riding [13]
2.8 Smart Tactile Gloves for Haptic Interaction, Communication and Rehabilitation [14]
3 Working/Methodology
3.1 Sensing and Analog Signal Condition Block
3.2 Continuous Force Monitoring Block
3.3 Haptic Feedback Block
4 Design of the Glove
5 Comparison
5.1 Smart Muscle Strength Assessment Glove for Rehabilitation Purposes [7]
5.2 Calibration and Evaluation of a Force Measurement Glove for Field-Based Monitoring of Manual Wheelchair Users [8]
5.3 A Fabricated Force Glove that Measures Hand Forces During Activities of Daily Living [9]
5.4 Wearable System with Embedded Force Sensors for Neurologic Rehabilitation Trainings [10]
5.5 Tactile Sensorized Glove for Force and Motion Sensing [11]
5.6 Smart Glove to Measure a Grip Force of the Workers [12]
5.7 Force-Sensing Glove System for Measurement of Hand Forces During Motorbike Riding [13]
6 Testing
7 Conclusion
References
Accident Detection in Surveillance Camera
1 Introduction
2 Literature Review
3 Problem Statement
4 Implementation
4.1 Vehicle Detection
4.2 Accident Detection
4.3 Alert System
5 Results
6 Conclusion
References
Wheeled Robots for Isolation Wards
1 Introduction
2 Literature Survey
3 Methodology
4 Technology Overview
5 Results and Discussion
6 Conclusion
References
A Survey on Various Crypto-steganography Techniques for Real-Time Images
1 Introduction
2 Application of Image Processing
3 Issues in Image Processing
3.1 Image Compression
3.2 Image Enhancement
3.3 Image Recognition
3.4 Image Visualization
3.5 Image Security
4 Application of Cryptography in Image Processing
5 Literature Survey
6 Conclusion
References
A Lightweight Image Cryptosystem for Multimedia Internet of Things
1 Introduction
2 Chaos-Based Encryption and Decryption
2.1 Chaos Maps
2.2 Chaos Maps for Encryption
2.3 Elliptic-Curve Cryptography Key Algorithm
3 Proposed Lightweight Image Encryption Algorithm
3.1 Key Generation Algorithm
3.2 Algorithm for Key Generation
3.3 Encryption Process
3.4 Confusion Process Utilising ACM
3.5 Diffusion Process Utilising Henon Map
3.6 Decryption Process
4 Results and Discussions
4.1 Input and Output Cases
4.2 Intensity Histogram Analysis
4.3 Entropy Analysis
4.4 Computational Complexity Analysis
4.5 Adjacent Pixel Auto Correlation Analysis
5 Conclusions
References
A Study on Parking Space Allocation and Road Edge Detection for Optimizing Road Traffic
1 Introduction
2 Study on Parking Space Allocation
3 Study on Road Edge Detection in Unfavourable Weather Conditions
4 Proposed Model
5 Conclusion
References
Human Physical Activities Based Calorie Burn Calculator Using LSTM
1 Introduction
2 Related Work
3 Methodology
3.1 Process Flowchart
3.2 Data Collection
3.3 Data Cleaning and Data Labeling
3.4 Patterns in sensor’s Data Recorded in the Activity
3.5 Proposed Methodology for Activity Recognition Using LSTM
3.6 Training of Model and Testing of Model
3.7 Calculating Calorie Deficit Using the Predicted Graph
4 Result and Conclusion
4.1 Accuracy of Model
4.2 Calorie-Meter on Andorid App
5 Future Directions
References
Alternate Tiny Encryption Algorithm: A Modified Tiny Encryption Algorithm for Improved Data Security
1 Introduction
2 Related Work
3 Proposed System
3.1 Key Generation
3.2 Subkey Generation
3.3 Encryption
3.4 Decryption
4 Results
5 Conclusion
References
Crystal Clear Analysis of Open–Source Automation Platforms
1 Introduction
2 Literature Summary of Open Source Software
2.1 OpenHAB
2.2 Home Assistant
2.3 Jeedom
2.4 IoBroker
2.5 Google Home Assistant
2.6 AGO Control
2.7 Domoticz
2.8 FHEM
2.9 Calaos
2.10 Pimatic
2.11 Homebridge.io
2.12 Node Red
2.13 IFTTT
3 Conclusion
4 Future Work
References
A Review Paper on Network Intrusion Detection System
1 Introduction
2 Architecture of IDS
3 Overview of ML Approaches
3.1 Supervised Learning
3.2 Unsupervised Learning
3.3 Reinforcement Learning
4 Data Reduction Approaches Used in IDS
4.1 Feature Selection
4.2 Feature Extraction
4.3 Clustering
5 Literature Survey
6 NIDS Datasets
7 Conclusion
References
ESP32 Based Irrigation System
1 Introduction
2 Literature Survey
3 Proposed System
3.1 ThingSpeak
3.2 Soil Moisture Sensor
3.3 Raindrop Sensor
3.4 DHT11 Sensor
3.5 Water Level Sensor
3.6 Methodology of the Proposed Work
4 Proposed System
4.1 Methodology of the Proposed Work
5 Conclusion
References
RFID (Radio Frequency Identification) Tag Collision Risk Mitigation Analysis and Avoidance
1 Introduction
2 Types of RFID Tags
3 Background
4 Types of Tag Collision
5 Risks of Tag Collision
6 Avoiding RFID Tag Collision
6.1 Dynamic Binary Tree Slot ALOHA (DBTSA)
6.2 Q Protocol
7 Comparison of ALOHA Based Anti-collision Protocols
8 Future of RFID (Radio Frequency Identification) Technology
9 Conclusion
References
BizGuru 1.0: Design and Development of a Mobile-Based Digital Marketing Guide for Elderly
1 Introduction
2 Related Work
2.1 Smartphone for Seniors (S4S) Project
2.2 Enhancing Islamic Knowledge via Short Messaging System (SMS)
2.3 eVideo Mobile
2.4 The SenApp Project
2.5 iPractice: Tablet-Based Home Practice Program in Aphasia Treatment
2.6 Reflection
3 Methodology
3.1 Analysis
3.2 Design
3.3 Development
3.4 Implementation
3.5 Evaluation
4 Conceptual Process Flow and Model of BizGuru
5 Findings and Discussion
5.1 Subject Matter Expert
5.2 Usability Evaluation
5.3 Related Theory and Methodological
6 Conclusion and Future Work
References
Development of Secure Cloud-Based Healthcare Management Using Optimized Elliptic Galois Cryptography
1 Introduction
2 Related Works
3 Proposed System
3.1 Architecture Overview
3.2 System Model
4 Results and Discussion
4.1 Performance Analysis of Proposed Mayfly Algorithm
4.2 Performance Analysis of Proposed Optimized EGC
5 Conclusion
References
A Review of Mobile Computation Offloading Techniques
1 Introduction
2 Mobile Task Offloading
3 Major Criteria in Offloading
4 Classification
5 Offloading Decision
6 Computation Offloading
7 Offloading Techniques
8 Conclusion
References
Study of the Impact of Sybil Attack in VANETs Using F2MD
1 Introduction
2 Literature Review on Sybil Attack
3 Sybıl Attack
4 Attacker Model
5 Sımulatıon
5.1 Evaluation
5.2 Performance Evaluation Metrics
6 Results
7 Conclusion
References
Aatmanirbhar Sanchar: Self-Sufficient Communications
1 Introduction
2 Motivation
3 Literature Review
4 Lacuna in the Existing System
5 Methodology
5.1 Joining a Chat Room
5.2 The Ephemeral Chat Room
5.3 Encryption Process
5.4 Transmitting a Message
5.5 Receiving a Message
5.6 Leaving the Chat Room
5.7 Experimental Environment Details
6 Applications
6.1 In Large Organizations as a Quick, Secure Chat Platform
6.2 Communications Where Security is of National Importance
6.3 Aatmanirbhar Samakraman: File Auto-Synchronization App
7 Results
7.1 Comparison of Results with Existing Systems
8 Future Scope
8.1 Converting the Application to a Decentralized WEB3 Application (See Fig. 15)
9 Conclusion
References
A Meta Heuristics SMO-SA Hybrid Approach for Resource Provisioning in Cloud Computing Framework
1 Introduction
1.1 Spider Monkey Optimization
1.2 Simulated Annealing
2 Related Work
3 Proposed Hybrid Algorithm (SMO-SA)
3.1 Mathematical Formulation
3.2 Algorithm of Proposed Hybrid Approach (SMO-SA)
4 Working Example of the Proposed Hybrid Algorithm (SMO-SA)
5 Results and Discussion
6 Conclusion
References
A Comprehensive Study of Automation Using a WebApp Tool for Robot Framework
1 Introduction
2 Associated Works
3 Proposed Work
3.1 Test Suite Creation
3.2 Test Case Execution
4 Result Analysis
5 Conclusion
References
Detection of Mirai and GAF-GYT Attack in Wireless Sensor Network
1 Introduction
2 Background
2.1 Mirai Attack
2.2 GAF-GYT Attack
3 Related Work
4 Proposed Method
4.1 Attack Detection System with F_classif Feature Selection
4.2 Normalization
4.3 Feature Extraction Using f_classif Function
5 Results and Discussion
5.1 Experimental Setup
5.2 Performance Analysis with Positive Measure Under GAF-GYT Attack
5.3 Performance Analysis with Negative Measure Under GAF-GYT Attack
5.4 Performance Analysis with Positive Measure Under Mirai Attack
5.5 Performance Analysis with Negative Measure Under Mirai Attack
6 Conclusion
References
A Brief Review of Network Forensics Process Models and a Proposed Systematic Model for Investigation
1 Introduction
2 Related Works
3 Systematic Networkt Forensic Investigation Model (SNFIM)
4 Classification of Internet Forensic Investigation Technologies
4.1 Intrusion Detection Systems of Internet Forensic Investigation
4.2 Traceback in Internet Forensic Investigation
4.3 Distributive Internet Forensic Investigation
4.4 Attack Graph of Internet Forensic Investigation
5 Methodology
5.1 Research Questions
5.2 Quality Assessment
5.3 Data Extraction and Data Synthesis
5.4 State-of-the-Art Internet Forensic Models
6 Comparative Analysis of State-of-Art Models
6.1 Intrusion Detection System in Network Forensic Analysis
6.2 Tracebacks in Network Forensic Analysis
6.3 Distributive Systems in Forensic Analysis
6.4 Attack Graphs in Forensic Analysis
7 Conclusion
References
IOT Based Solution for Effective Social Distancing and Contact Tracing for COVID-19 Prevention
1 Introduction
2 Related Work
3 Proposed Method
3.1 IR Sensor
3.2 Ultrasonic Sensor
4 Social Distancing Algorithm
4.1 Algorithm: Social Distancing Measurement
4.2 Temperature Checking
5 Result and Analysis
6 Conclusions
References
Design and Implementation of Highly Secured Nano AES Cryptographic Algorithm for Internet of Things
1 Introduction
1.1 Internet-of-Things
1.2 Crypto System
1.3 Advanced Encryption Standard (AES)
1.4 Deoxyribonucleic Acid (DNA)
2 Literature Survey
3 Design Implementation
4 Result and Discussions
5 Conclusion
References
Convergence of Communication Technologies with Internet of Things
1 Introduction
2 Literature Survey
3 RFID and IoT
4 Wireless Sensor Network
5 Connecting WSN with IoT
6 Connecting MANET’s to IoT
7 MANET-IoT Integration Protocol
8 Results and Discussion
9 Conclusion
References
Chatbots: A Survey of the Technology
1 Introduction
1.1 Chatbot Statistics
1.2 History of Chatbots
2 Types of Chatbots
2.1 Knowledge Domain
2.2 Service Provided
2.3 Goals
2.4 Method for Processing Input and Producing Responses
3 Characteristics of Chatbot
3.1 Conversational Development
3.2 Autonomous Judgement
3.3 Omni Channel
3.4 Accessible
3.5 Free to Explore
3.6 Encourage Analytics
3.7 Secure
3.8 Allows Upgrades
3.9 Pre-trained
3.10 Emotionally Intelligent
3.11 Integrates with CRM
3.12 Sentences Are to the Point
3.13 A Distinct Personality
3.14 Handles the Unexpected
3.15 Saves Money and Time
4 Chatbot Approaches, Tools and Architecture
4.1 Approaches
4.2 Chatbot Tools
4.3 Chatbot Architecture
5 Metrics for Evaluation of Chatbot
5.1 Chatbot Activity
5.2 Chatbot Interaction Rate
5.3 Chatbot Session Length
5.4 Chatbot Goal Completion Rate
5.5 Chatbot Retention Rate
5.6 Chatbot Non-response Rate
5.7 Questions Per Conversation
6 Comparison of Chatbots
7 Chatbots in Healthcare
7.1 Can We Trust Health in the Hands of Chatbots?
7.2 Benefits of Chatbot in Healthcare
8 Open Challenges Related to Chatbot
9 Proposed System
10 Conclusion
References
An Improved Machine Learning Algorithm for Crash Severity and Fatality Insight in VANET Network
1 Introduction
2 Related Work
3 Crash Severity and Fatality Insight
3.1 Accident Location Analysis
3.2 Accident Happened Due to Lighting Conditions
3.3 Accident Lane Analysis
3.4 Analysis on Accident Happened Due to Weather
3.5 Accident Caused by Over Speed
4 Performance Analysis and Result
5 Conclusion
References
Network Monitoring of Cyber Physical System
1 Introduction
2 Review of Literature
3 Basic Notions
3.1 Networking
3.2 Networking
4 System Modules
5 Proposed Methodology
6 Simulated Results
7 Importance of the Monitoring Tool
8 Conclusion
References
Impact of Security Attacks on Congestion in Wireless Sensor Networks
1 Introduction
2 Related Works
3 Attacks that Contribute to Congestion Occurrence
3.1 Packet Drop Method
3.2 Flooding Method
3.3 Delay Method
3.4 False Route Method
4 Experimental Results
4.1 Effect of Attacks in Different Network Scenarios
4.2 Effect of Attacks on Important Network Parameters
5 Conclusion
References
IoT Weather Forecasting Using Ridge Regression Model
1 Introduction
2 Hardware Requirements
3 Software Requirements
4 Working and Design
5 Literature Survey
6 Equations
7 Result Screenshots
8 Conclusion
9 Future Enhancement
References
Automated Cloud Monitoring Solution: Review
1 Introduction
2 Motivation to Design
3 Architecture of Monitoring System
4 Results and Performance Analysis
5 Advancement
6 Comparison of Cloud Monitoring Solutions
7 Conclusion
References
A Secured Framework Against DDoS Attack in Wireless Networks
1 Introduction
2 Related Works
3 Proposed Model
3.1 Phase I: DDoS Attack Implementation
3.2 Phase II: Attack Detection
3.3 Phase III: Attack Elimination
3.4 Phase IV: Authentication
4 Result Analysis
5 Conclusion
References
Anomaly Based Intrusion Detection System Using Rule Based Genetic Algorithm
1 Introduction
2 Literature Survey
3 Methodology of Proposed System
3.1 Ensemble Classifier
3.2 Anomaly Detection: Genetic Algorithm
4 Results and Discussions
4.1 Performance Using UNSW NB15
5 Conclusion
References
Hybrid Learning Approach for E-mail Spam Detection and Classification
1 Introduction
2 Related Works
2.1 Machine Learning Techniques
2.2 Deep Learning Techniques
3 Proposed Work
3.1 Detecting Spam E-mail Based on Text Content
3.2 Detecting Spam E-mail Based on URL
3.3 Detecting Spam E-mail Based on Message Id
3.4 Model Aggregation
4 Result Analysis
5 Conclusion
References
Smart Solid Waste Management System Using IoT Technology: Comparative Analysis, Gaps, and Challenges
1 Introduction
1.1 Paper Organization
2 Analysis of Previous Works
2.1 IoT-Enabled Smart Bin
2.2 Smart Waste Management in the Context of Smart Cities
2.3 Smart Garbage Monitoring and Route Optimizing System
2.4 Comparative Analysis and Research Gaps
3 Challenges and Issues for Smart Waste Management System
3.1 System Segregation at the Collection Level
3.2 Optimal Collection Route
3.3 Communication Technology
3.4 User-Friendly System
3.5 Lifetime Maximization of IoT Devices
4 Critical Discussion
5 Conclusion
References
HLWEA-IOT: Hybrid Lightweight Encryption Algorithm Based Secure Data Transmission in IoT-MQTT Networks
1 Introduction
2 Related Works
2.1 Problem Statement
3 Methodology
3.1 Framework of SAEMS
3.2 Hybrid Lightweight Encryption Algorithm (HLWEA)
4 Results and Discussions
5 Conclusion
References
A Practical Approach for Crop Insect Classification and Detection Using Machine Learning
1 Introduction
2 Literature Review
2.1 Noteworthy Contribution
3 Literature Summary
4 Objectives
4.1 To Review and Understand the Morphology of the Insects
4.2 To Detect and Classify the Pest Based on Shape and Size Features, Using ML Techniques
4.3 To Study Various Machine Learning and Deep Learning Algorithm for Insect Classification and Identification
4.4 To Develop the Hybrid Machine-Learning Algorithm for the Classification and Detection of Insects
5 Recent Study
6 Proposed Work
6.1 Dataset and Methods
6.2 Image Preprocessing
6.3 Features Extracted and Method of Extraction
6.4 Methodology
6.5 Filter Size for Each Layer
6.6 Classification Accuracy
6.7 Computational Time
7 Conclusion
References
Attendance Portal Using Face and Speaker Recognition
1 Introduction
2 Literature Survey
3 Methodology
3.1 Data Gathering and Preprocessing
3.2 Model Building and Training
3.3 Interface Development
4 Results
5 Conclusion
References
Blockchain-Enabled Network for 6G Wireless Communication Systems
1 Introduction
2 6G General Specification
2.1 Higher Data Rate
2.2 Massive Number of Users
2.3 Security Requirements
3 Blockchain Terminology
3.1 Consensus Protocols
3.2 Blockchain Classification
3.3 Smart Contracts
3.4 Blockchain Features
4 Blockchain Solutions for 6G
4.1 Blockchain-Enabled 6G Technology
4.2 Blockchain-Enabled 6G Use Cases
5 Challenges
5.1 Storage
5.2 Delay
5.3 Security Risks
5.4 Scalability
6 Conclusion
References
Machine Learning Based Automated Disaster Message Classification System Using Linear SVC Algorithm
1 Introduction
2 Literature Survey
3 Proposed System of Disaster Prediction
4 Method of the Proposed System
5 Results
6 Conclusion
References
Intelligent Healthcare System
1 Introduction
2 Literature Review
3 Methodology
4 Experimental Result and Discussion
5 Conclusion
6 Future Scope
References
Intelligent Predictive Maintenance for Industrial Internet of Things (IIoT) Using Machine Learning Approach
1 Introduction
2 Related Work
3 Proposed Work Methodology
3.1 Layer Wise Application Oriented IIoT Architecture
3.2 Data Acquisition Using Experimental Set Up
3.3 Data Analytics
3.4 Data Analytics Algorithm Steps
4 Results Analysis and Discussion
5 Conclusion
References

Citation preview

Engineering Cyber-Physical Systems and Critical Infrastructures 3

Jude Hemanth · Danilo Pelusi · Joy Iong-Zong Chen Editors

Intelligent Cyber Physical Systems and Internet of Things ICoICI 2022

Engineering Cyber-Physical Systems and Critical Infrastructures Volume 3

Series Editor Fatos Xhafa , Departament de Ciències de la Computació, Technical University of Catalonia, Barcelona, Spain

The aim of this book series is to present state of the art studies, research and best engineering practices, real-world applications and real-world case studies for the risks, security, and reliability of critical infrastructure systems and Cyber-Physical Systems. Volumes of this book series will cover modelling, analysis, frameworks, digital twin simulations of risks, failures and vulnerabilities of cyber critical infrastructures as well as will provide ICT approaches to ensure protection and avoid disruption of vital fields such as economy, utility supplies networks, telecommunications, transports, etc. in the everyday life of citizens. The intertwine of cyber and real nature of critical infrastructures will be analyzed and challenges of risks, security, and reliability of critical infrastructure systems will be revealed. Computational intelligence provided by sensing and processing through the whole spectrum of Cloud-to-thing continuum technologies will be the basis for real-time detection of risks, threats, anomalies, etc. in cyber critical infrastructures and will prompt for human and automated protection actions. Finally, studies and recommendations to policy makers, managers, local and governmental administrations and global international organizations will be sought.

Jude Hemanth · Danilo Pelusi · Joy Iong-Zong Chen Editors

Intelligent Cyber Physical Systems and Internet of Things ICoICI 2022

Editors Jude Hemanth Department of Electronics and Communication Engineering Karunya Institute of Technology and Sciences Coimbatore, Tamil Nadu, India

Danilo Pelusi Faculty of Communication Sciences University of Teramo Teramo, Italy

Joy Iong-Zong Chen Department of Electrical Engineering Da-Yeh University Dacun, Changhua, Taiwan

ISSN 2731-5002 ISSN 2731-5010 (electronic) Engineering Cyber-Physical Systems and Critical Infrastructures ISBN 978-3-031-18496-3 ISBN 978-3-031-18497-0 (eBook) https://doi.org/10.1007/978-3-031-18497-0 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

We are honored to dedicate the proceedings of ICoICI 2022 to all the participants and editors of ICoICI 2022.

Preface

It is with deep satisfaction that I write this preface to the proceedings of the ICoICI 2022 held in JCT College of Engineering and Technology, Coimbatore, Tamil Nadu, India, during August 11–12, 2022. This conference proceedings volume contains the written versions of most of the contributions presented during the conference of ICoICI 2022. The conference provided a setting for discussing recent developments in a wide variety of topics including Cyber-Physical Systems, Data Communication, Computer Networking, Communicational Technologies, Cryptography, Big Data, Cloud Computing, IoT, and Healthcare Informatics. The conference has been a good opportunity for participants coming from various destinations to present and discuss topics in their respective research areas. ICoICI 2022 conference tends to collect the latest research results and applications on Intelligent Cyber Physical Systems and Internet of Things. It includes a selection of 65 papers from 302 papers submitted to the conference from universities and industries all over the world. All of accepted papers were subjected to strict peerreviewing by 2–4 expert referees. The papers have been selected for this volume because of quality and the relevance to the conference. ICoICI 2022 would like to express our sincere appreciation to all authors for their contributions to this book. We would like to extend our thanks to all the referees for their constructive comments on all papers, especially, we would like to thank to committee members for their hard working. Finally, we would like to thank the Springer publications for producing this volume.

Coimbatore, India Teramo, Italy Dacun, Taiwan

Guest Editors—ICoICI 2022 Prof. Dr. Jude Hemanth Dr. Danilo Pelusi Prof. Dr. Joy Iong-Zong Chen

vii

Contents

Term Frequency Tokenization for Fake News Detection . . . . . . . . . . . . . . . Pallavi Suresh, Abhishek Shettigar, M. Karunavathi, Ajith, and M. G. Ramanath Kini

1

Aquaculture Monitoring System Using Internet of Things . . . . . . . . . . . . . G. V. R. Kameshwar Rao, T. J. Dhivya Shrilaa, I. Akash, and G. Gugapriya

11

A Comprehensive Study and Implementation of Memory Malware Analysis with Its Application for the Case Study of CRIDEX . . . . . . . . . . Digvijay Singh and Rajesh Yadav IoT Based Anti Poaching of Trees and Protection of Forest . . . . . . . . . . . . E. V. Kameswararao, M. Jaya Shankar, T. V. Sai Lokesh, and E. Terence Artificial Intelligence Based Efficient Activity Recognition with Real Time Implementation for ATM Security . . . . . . . . . . . . . . . . . . . . S. Srinivasan, AL. Vallikannu, Adapa Sankar Ganesh, Iragamreddy Raj Kumar, and Beereddy Venu Gopal Terror Attack Classification with the Application of Orange Data Mining Tool and Neo4j Sandbox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ankit Raj, Suchitra A. Khoje, and Sagar Bhilaji Shinde Multipurpose IoT Based Camera Using Deep Learning . . . . . . . . . . . . . . . Urvashi Dube, Sudhish Subramaniam, and G. Sumathi

31 45

57

69 85

Dr. Watson AI Based Healthcare Technology Project . . . . . . . . . . . . . . . . . . 101 N. Suresh Kumar, S. Ganesh Karthick, K. P. Aswin Kumar, S. Balaji, and T. Nandha Sastha Empirical and Statistical Comparison of RSA and El-Gamal in Terms of Time Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 Ankita Kumari, Prashant Pranav, Sandip Dutta, and Soubhik Chakraborty

ix

x

Contents

IoT Communication to Capture and Store Data to Thingspeak Cloud Using NodeMCU and Ultrasonic Sensor . . . . . . . . . . . . . . . . . . . . . . . 121 Priya J. Payyappilly and Shweta Dour A Comprehensive Study on Cloud Computing: Architecture, Load Balancing, Task Scheduling and Meta-Heuristic Optimization . . . . . . . . . 137 Shruti Tiwari and Chinmay Bhatt Balancing Exploration and Exploitation in Nature Inspired Computing Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 K. Praveen Kumar, Sangeetha Singarapu, Mounika Singarapu, and Swaroop Rakesh Karra Blockchain Based Secure, Efficient, and Scalable Platform for the Organ Donation Process of Healthcare Industry . . . . . . . . . . . . . . . 173 Keyur Parmar, Vadlapudi Devanand Kumar, Neduri Leela Prasanth, Pranoppal, Kasa Charan Teja, Shriniwas Patil, and Kaushal A. Shah Image Enhancement in Frequency Domain Fingerprint Detection and Matching Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185 Suhasini S. Goilkar and Shashikant S. Goilkar Developing Machine Learning Based Mobile App for Agriculture Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 R. Dhivya and N. Shanmugapriya Attack Detection in IoT Using Machine Learning—A Survey . . . . . . . . . . 211 Saeed Ali Haifa Ali and J. Vakula Rani An Extensive Study on Logic Emerging IoT Adiabatic Techniques for Low-Power Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229 T. Vijayalakshmi and J. Selvakumar A Critical Review of Agri-Food Supply Management with Traceability and Transparency Using Blockchain Technology . . . . . 239 Sanket Araballi and P. Devaki Face-Anti-spoofing Based on Liveness Detection . . . . . . . . . . . . . . . . . . . . . . 251 Shivani Mangal and Khushboo Agarwal PDR Analysis and Network Optimization of Routing Protocols for Edge Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265 Archana Ratnaparkhi, Radhika Purandare, Gauri Ghule, Shraddha Habbu, Arti Bang, and Pallavi Deshpande Privacy Threat Reduction Using Modified Multi-line Code Generation Algorithm (MMLCGA) for Cancelable Biometric Technique (CBT) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275 Pramod D. Ganjewar, Sanjeev J. Wagh, and Aarti L. Gilbile

Contents

xi

Systematic Literature Review—IoT-Based Supply Chain Management in Industry 4.0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291 Sreeparnesh Sharma Sivadevuni and Sathish Kumar Ravichandran A Review on Urban Flood Management Techniques for the Smart City and Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303 Anil Mahadeo Hingmire and Pawan R. Bhaladhare Application of Distributed Constraint Optimization Technique for Privacy Preservation in Cyber-Physical Systems . . . . . . . . . . . . . . . . . . 319 Manas Kumar Yogi and A. S. N. Chakravarthy Grip Assisting Glove for Charcot-Marie-Tooth Patients . . . . . . . . . . . . . . . 329 Varun Sarathchandran, Jason Vincent, Juel Mathais George, Polu Sathwik Reddy, and R. Ambika Accident Detection in Surveillance Camera . . . . . . . . . . . . . . . . . . . . . . . . . . 345 A. P. Adil, M. G. Anandhu, Jeovan Elsa Joy, Twinkle S. Karethara, S. Anjali, and B. R. Poorna Wheeled Robots for Isolation Wards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355 U. Sahana and N. Rajesh A Survey on Various Crypto-steganography Techniques for Real-Time Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365 R. Tanya Bindu and T. Kavitha A Lightweight Image Cryptosystem for Multimedia Internet of Things . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375 V. Panchami, Arjun Rajasekharan, and Mahima Mary Mathews A Study on Parking Space Allocation and Road Edge Detection for Optimizing Road Traffic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393 H. Varun Chand, Seema Sabharwal, Anil Carie, and S. Arun Kumar Human Physical Activities Based Calorie Burn Calculator Using LSTM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405 Jadhav Kalpesh, Jadhav Rushikesh, Kalbande Swaraj, Katta Rohan, and Rakhi Bharadwaj Alternate Tiny Encryption Algorithm: A Modified Tiny Encryption Algorithm for Improved Data Security . . . . . . . . . . . . . . . . . . . 425 Mehak Gupta, Nimit Agrawal, and Manas Ranjan Prusty Crystal Clear Analysis of Open–Source Automation Platforms . . . . . . . . . 437 Kiran Jadhav, Mangesh Nikose, and Sagar Shinde A Review Paper on Network Intrusion Detection System . . . . . . . . . . . . . . 453 Nongmeikapam Thoiba Singh and Raman Chadha

xii

Contents

ESP32 Based Irrigation System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465 M. Koteswara Rao, M. Satish Kumar, M. Jaijaivenkataramana, and Ch. Sai Sowjanya RFID (Radio Frequency Identification) Tag Collision Risk Mitigation Analysis and Avoidance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475 Aditya Sukhwal, Gourab Kundu, and Chandrani Chakravorty BizGuru 1.0: Design and Development of a Mobile-Based Digital Marketing Guide for Elderly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487 Ahmad Sofian Shminan, Nur Zulaikha Mohamed Aziyen, Lee Jun Choi, and Merikan Aren Development of Secure Cloud-Based Healthcare Management Using Optimized Elliptic Galois Cryptography . . . . . . . . . . . . . . . . . . . . . . . 505 V. Gokula Krishnan, D. Siva, S. MuthuSelvi, T. A. Mohana Prakash, P. A. Abdul Saleem, and S. Mary Rexcy Asha A Review of Mobile Computation Offloading Techniques . . . . . . . . . . . . . . 519 M. Jyothirmai, Kesavan Gopal, and M. Sailaja Study of the Impact of Sybil Attack in VANETs Using F2MD . . . . . . . . . . 533 T. Pavithra, B. S. Nagabhushana, and Suchismitha Das Aatmanirbhar Sanchar: Self-Sufficient Communications . . . . . . . . . . . . . . 545 Jay Jhaveri, Abhay Gupta, Prem Chhabria, Neeraj Ochani, Sharmila Sengupta, Mrs. Sunita Suralkar, and Shashi Dugad A Meta Heuristics SMO-SA Hybrid Approach for Resource Provisioning in Cloud Computing Framework . . . . . . . . . . . . . . . . . . . . . . . 563 Archana and Narander Kumar A Comprehensive Study of Automation Using a WebApp Tool for Robot Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577 N. Alok Chakravarthy and Usha Padma Detection of Mirai and GAF-GYT Attack in Wireless Sensor Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 587 Hanjabam Saratchandra Sharma, Moirangthem Marjit Singh, and Arindam Sarkar A Brief Review of Network Forensics Process Models and a Proposed Systematic Model for Investigation . . . . . . . . . . . . . . . . . . . 599 Merly Thomas and Bandu Meshram IOT Based Solution for Effective Social Distancing and Contact Tracing for COVID-19 Prevention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 629 S. Kanakaprabha, P. Arulprakash, V. Priyanka, Vineetha Varghese, and A. Sureshkumar

Contents

xiii

Design and Implementation of Highly Secured Nano AES Cryptographic Algorithm for Internet of Things . . . . . . . . . . . . . . . . . . . . . . 645 E. Roopa and Yasha Jyothi M. Shirur Convergence of Communication Technologies with Internet of Things . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 659 V. Dankan Gowda, Suma Sira Jacob, Naziya Hussain, R. Chennappan, and D. T. Sakhare Chatbots: A Survey of the Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 671 Hrithika Singh, Asmita Bhangare, Rashmi Singh, Shubhangi Zope, and Pallavi Saindane An Improved Machine Learning Algorithm for Crash Severity and Fatality Insight in VANET Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693 S. Bharathi and P. Durgadevi Network Monitoring of Cyber Physical System . . . . . . . . . . . . . . . . . . . . . . . 705 Mayank Srivastava, Aman Maurya, Utkarsh Sharma, and Shikha Srivastava Impact of Security Attacks on Congestion in Wireless Sensor Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 721 Divya Pandey and Vandana Kushwaha IoT Weather Forecasting Using Ridge Regression Model . . . . . . . . . . . . . . 733 Karthik G. Dath, K. E. Krishnaprasad, T. S. Pushpa, and K. P. Shailaja Automated Cloud Monitoring Solution: Review . . . . . . . . . . . . . . . . . . . . . . 747 Ishwari Deshmukh and Jayshri D. Pagare A Secured Framework Against DDoS Attack in Wireless Networks . . . . 757 O. K. Vismaya, Ajay Kumar, Arya Paul, and Albins Paul Anomaly Based Intrusion Detection System Using Rule Based Genetic Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 769 Shraddha R. Khonde Hybrid Learning Approach for E-mail Spam Detection and Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 781 Rimitha Shajahan and P. L. Lekshmy Smart Solid Waste Management System Using IoT Technology: Comparative Analysis, Gaps, and Challenges . . . . . . . . . . . . . . . . . . . . . . . . 795 Meenakshi Shruti Pal and Munish Bhatia HLWEA-IOT: Hybrid Lightweight Encryption Algorithm Based Secure Data Transmission in IoT-MQTT Networks . . . . . . . . . . . . . . . . . . . 813 S. Hariprasad, T. Deepa, and N. Bharathiraja

xiv

Contents

A Practical Approach for Crop Insect Classification and Detection Using Machine Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 825 Ravindra Yadav and Anita Seth Attendance Portal Using Face and Speaker Recognition . . . . . . . . . . . . . . . 845 Sahil Sharma, Shivam Prajapati, Merin Meleet, and B. S. Rekha Blockchain-Enabled Network for 6G Wireless Communication Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 857 Nazanin Moosavi and Hamed Taherdoost Machine Learning Based Automated Disaster Message Classification System Using Linear SVC Algorithm . . . . . . . . . . . . . . . . . . . 869 N. Merrin Prasanna, S. Raja Mohan, K. Vishnu Vardhan Reddy, B. Sai Kumar, C. Guru Babu, and P. Priya Intelligent Healthcare System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 881 M. Senthamil Selvi, K. Abinaya, N. Jemy Sharon, and R. Lakshmi Pooja Intelligent Predictive Maintenance for Industrial Internet of Things (IIoT) Using Machine Learning Approach . . . . . . . . . . . . . . . . . . 897 Umesh W. Hore and D. G. Wakde

Term Frequency Tokenization for Fake News Detection Pallavi Suresh, Abhishek Shettigar, M. Karunavathi, Ajith, and M. G. Ramanath Kini

Abstract In today’s world, when the internet is pervasive, everyone gets news from a variety of online sources. As the use of social media platforms has grown, news has travelled quickly among thousands of people in a very less duration. The propagation has been far reaching for the fake news generation in repercussions, from altering election outcomes in support of specific politicians, creating prejudiced viewpoints. Furthermore, spammers use appealing news headlines to make cash through click-bait adverts. In today’s world knowingly or unknowingly fake news spreads around the world through internet. This has a great impact on the people who blindly believe whatever the internet provides. Hence, fake news identification has become a new study subject that is attracting a lot of attention. However, due to a lack of resources, such as datasets and processing and analysis procedures, it encounters several difficulties. This research uses a non-probabilistic machine learning models of computational prototypes to address this problem. Furthermore, the comparison of Term Frequency-Inverse Document Frequency (TF-IDF) is done, for the purpose of determining the best vectorizer used for detecting fake news. In order to raise the accuracy, stop words of English are used. To predict bogus news, a Support Vector Machine (SVM) classifier is deployed. According to the simulation data, the SVM and the TF-IDF produce results with high accuracy. Keywords Fake news detection · TF-IDF · SVM · Machine learning

P. Suresh · A. Shettigar (B) · M. Karunavathi · Ajith · M. G. Ramanath Kini Department of Electronics and Communication Engineering, Mangalore Institute of Technology and Engineering, Moodabidri, Mangalore, Karnataka, India e-mail: [email protected] M. G. Ramanath Kini e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_1

1

2

P. Suresh et al.

1 Introduction As more people spend their time communicating in social media platforms through online, they are consuming more news and seeking out news through social media instead of referring to the other regional news organization. Many people believe false news influenced the 2016 presidential election campaign in some countries. Therefore, the word ‘fake’ news has entered the common language as a result of this election [1]. In today’s world fake news has become a major problem in case of family, friends, and coworkers, and the core reason for this is social media. People disseminates a large area of news on specific topics, for example, the issues about the government, famous properties, etc. through internet. Bunch of fake rumors about the Covid-19 has spread through the internet between 2020 of January and 2020 of May. Using the modification of text, during the pandemic, people had spread false information. The goal of such kind of people is to spread false news to further incite political agenda and religious conflict. The fake news and rumors which comes from the internet cause problems in every region, including the health, agriculture, the private sector, school and colleges, industries, banks, and so on. Fake news is created to deceive the public in a developed and verified false manner [2, 3]. Micro-blog, Instagram, blogs, and Facebook are examples of social media platforms that can speed up the information among users which accelerate the distribution to every corner of the world. For the transformation of the information and for the social interaction, the social media has been used as the primary platform due to its user- friendliness, low cost, and rapid rate. Fake news can take the visible shape of images, texts, audio, or video [1]. A fakester manipulates information in order to mislead the public. The spread of the data via web-based networking media is done from the clients without prior check. A proper methodology for the detection of fake news employs the following techniques: • Text preprocessing: To remove the stop words along with special characters, steaming and analyzing the text. • Text encoding: By the help of TF-IDF (Term Frequency- Inverse Document Frequency) • Characteristic extraction: For the accurate identification of fake information, including source from news and the author, the date, and with emotion conveyed through the text to be the news features. • Support Vector Machine: An improvised machine learning algorithm which can classify new data.

Term Frequency Tokenization for Fake News Detection

3

2 Literature Review To select the best individual model, researchers have gone through several models about machine learning from various datasets. Paper [4] used SVM for the purpose of classifier and TF-IDF bunch with texts along with n-grams for technique of feature extraction. In addition, they executed a dataset which contains both the true and fake news, to be trained for the upcoming model. Based on the test which are done through different experiments, the TF-IDF algorithm outshone among other various intelligent classification algorithms. In paper [5], to train the twenty-three datasets from the supervised artificial intelligence algorithms, three datasets were used. Except recall, the algorithm of decision tree surpassed various smart performance matrices using classification algorithms, according to their results obtained. In article [6], with the help of set that include the feature about the article structure encapsulation, readability, and title-to-body similarity, they chose six machine learning algorithms for the experiment such as, AB (AdaBoost), DT (Decision Tree), KNN (K-Nearest Neighbour), RF (Random Forest), SVM (Support Vector Machine), and XGB (XGBoost). The confusion matrix model with other evaluation metrics applied, quantify a classifier’s performance. Given the experiment’s structure, the best results were produced from the support vector machine classifier. Some researchers have developed extraordinary neural networks for detecting fake news. Article [7] developed a system with the methodology in order to point out fake news found in web articles by the help of machine learning, and natural language processing. For the generation of feature vector in the work proposed, various methods which features the count vector TF-IDF, with the collection of words were applied. To identify news as fake or as real, seven different machine learning classification algorithms were trained, especially in comparison in terms of accuracy, F1 Score, recall, and precision, before they chose the one which is best for building a system which have to predict a news as fake or real. Machine learning algorithms were used on different datasets from different sources in [8]. They also included each model’s analysis results. The difficult task of detecting fake news can be simplified by using the right models and tools. Some researchers have attempted to combat the wide spread of fake news on different social media sources. A feature of new set to predict quality from the current situations and characteristics of false information through detection of fake news was proposed in [9]. It shows intriguing findings about utility, also the significance about detecting false news features. Lastly, discussion about how to use fake news detection has been practiced, highlighting challenges as well as opportunities. Among those who contributed to this work is [10], that considered automating shaped news differentiating proof in Twitter datasets using a model for understanding mass produced news messages through post on the Twitter. Following that, they independently performed a similarity check within the five Machine Learning algorithms, such as the Support Vector Machine, Naive Bayes Method, Logistic Regression, and Recurrent Neural Network models, for the demonstration of the dataset to yield better result. The results revealed that SVM with Naive Bayes classifiers outperformed other

4

P. Suresh et al.

calculations. The theory-driven model was applied in [2] for the detection of the fake news from extracting false information and also clicks bait characteristics of news article text and using models of machine learning which includes SVM, Random Forest, XGBoost, Naive Bayes, and Logistic Regression. Some of the researchers have also made use of other news features to improve their results. A model was estimated in [11] to distinguish the fake news from news articles intuitively. It has proposed an unseen feature consisting of the set of classifiers for the purpose of machine learning. A dataset used in the experiment is a collection of both datasets containing balanced amounts of both true and fake news about politics. The linguistic/stylometric features, and bunch containing texts TF, and a BOW TFIDF vector were extracted from the dataset’s text fields, and then various machine learning models were applied, which include methods of bagging and boosting, for the purpose of obtaining the greater probability of accuracy. Purevdagva et al. [12] used an automated framework to detect phony political speech. It extracted the features of statements of political speech and their metadata using various classification methods, location, with the speech subject, writer’s credibility, writer’s profile, and also information of speech context. Machine learning model should be trained containing automatic selection, using features with parameterization. For a good amount of detection accuracy on the “Liar” dataset a SVM with trained model was achieved. The results which are evaluated show that the framework has been effective in detecting phony political speeches. While the other researchers have concluded deep learning framework by themselves and received excellent results on their datasets. The precision accuracy, F1score, and recall, evaluation measures, and also a proper version of McNemar’s test was performed in [13], to check if the model performance have different output. Then, on the ISOT and KDnugget datasets, a novel stacking model of novel received accuracy for training and testing with 99.94% and 96.05%, respectively. Furthermore, when compared to baseline methods, the proposed method performed well. As a result, they strongly recommend it for detecting fake news. Paper [14] evaluated and compared various approaches to mitigate this issue, along with approaching several traditional machine learning approaches, like Naive Bayes, also other famous deep learning algorithms like RNN and hybrid CNN. The comparison was made not only between traditional and deep learning algorithms, but also between traditional and non-traditional methods. The groundwork for deciding on machine learning or deep learning procedures used to solve problems that bring out a balance between accuracy and light weightlessness. Manzoor et al. [15] conducted a study on other machine learning approaches for detecting fake and made out news [16]. Limitations of these kinds of approaches and improvisation through the implementation of deep learning are also discussed.

3 Methodology The flow chart of the methodology is shown in Fig. 1.

Term Frequency Tokenization for Fake News Detection Fig. 1 Process of machine learning classification

5

Data Set Pre-processing the Data Feature Extrac on Training the Classifier Opining Classifier Fake Opinion

True Opinion

A. Dataset The dataset for the project has been selected from the site Kaggle, which included the data from several websites which included political news, government news, left news, middle east news, and US news, which has around 1 post recorded for 30 days. The main features consist of the title, text about the news body, subject about the news, authorized date, and the label, which included the training dataset of 18,525 posts and around 8980 for testing. B. Pre-processing Before the data are fed through machine learning algorithms, the text must be preprocessed using methodologies such as converting all letters from the document to the lowercase, removing stop words, tokenization, sentence fragmentation, and punctuation, and removing accent marks. These operations can greatly assist in choosing suitable terms for improving the performance of the model. Two datasets are derived through real-world news articles. Therefore, cleaning up data by removing the stop words is performed. To finish the sentence structure, stop words are commonly applied in the English sentences, but these have little significance in terms of showing an individual’s attitudes. As a result, they are removed from all experiments to prevent generating large unwanted noise. Further it is tokenized with TF-IDF by cleaning the data containing text. • Term Frequency-Inverse Document Frequency (TF-IDF) TF is a prevalent method used for tokenization which calculates in order to validate the proposed model by counting the set containing text available in the document. Every file is represented using a sequence containing the word that counts, which is

6

P. Suresh et al.

done using TF technique. Later, for all vectors, the sum of elements will be unique, converting word counts into accuracy. Consider ‘D’ as a corpus, and a document is represented as ‘d’. Assume ‘w’ as word in document and nw (d) will be the count of w in the data. As a result, dimensions of ‘d’ is denoted as:  |d| = nw (d) wed

The normalized TF used in the word ‘w’ in source ‘d’ is as follows: T F(w)d =

nw |d|

In the machine learning experiments, TF-IDF is applied to convert the data in the form of vectors. The TF-IDF is a type of weighing metrics that frequently comes in hand for classification of text problems. This also has to gain a score to each term in a document, indicating the importance of the term. The significance of a term in this method grows in proportion to its frequency containing dataset. Let ‘D’ denote a corpus, that is, a collection of articles from the news. In this, ‘d’ represents an article made up of the words ‘w’. Using the equation below, mathematical calculation of IDF is done.   |D| I D F(w) D = 1 + log ({|d| : D|w ∈ d|}) The term TF-IDF used in the word (w) in relation for the document (d) and corpus (D) is known by the following. TF − IDF(w)d,D = TF(w)d × IDF(w)D C. Classification: support vector machine (SVM) A SVM Classifier is a guided technique machine learning which helps to tackle different classification along with problem related to regression. This algorithm is mainly used for solving difficulty on categorization. Every data point in n-dimension is plotted as a point (where ‘n’ is considered as the number that the features contain), considering that value in every feature have its amount with particular coordinate present in SVM algorithm. Later, it should be accomplished for the classification to distinguish the data by locating the hyper-plane as shown in Fig. 2.

4 Results and Discussion In this work, first all letters from the document are converted to the lowercase, stop words are removed, sentence fragmentation, and punctuation, and removing

Term Frequency Tokenization for Fake News Detection

Maximum Margin

7

Posi ve Hyperplane

Maximum Margin Hyperplane

Nega ve Hyperplane

Support vector

Fig. 2 Support vector machine classification

accent marks are performed. Then, TF-IDF method is utilized for tokenization to obtain the text representation. After that, text representation future is trained by a machine learning models i.e., SVM. The evolution of accuracy percentage belonging on features is by testing on the training data. The dataset is explored using Google Colab, a free cloud service backed by Google. In this study, 80% of the datasets are utilized to train a classifier, while the other 20% of the datasets are used to test the classification model. The confusion matrix data collected at a classifier are summarized in Table 1. 1. False Positives is represented by FP, True Positives is represented by TP, False Negatives is represented by FN and True Negatives is represented by TN, which are the four metrics in the confusion matrix that can be used to evaluate a classifier’s performance. When the data are examined, the SVM classifier holds a minimal error rate of 28 articles for false negatives and 20 articles for false positives. Furthermore, the SVM classifier provides the most accurate classification of false at 4619 articles as TN and real at 4313 articles as TP. The equations for the contents in the Table 2 are: Precision :

4313 TP = × 100 = 99.53% (T P + F P) (4313 + 20)

Table 1 Confusion matrix results by classifier predicted label

True label

Predicted label TN

FP

4619

20

FN

TP

28

4313

8

P. Suresh et al.

Table 2 Performance results of precision, accuracy, recall and F-measure Classifier

Precision

Accuracy

Recall

F-Measure

SVM

99.3%

99.4%

99.4%

99.4%

Accuracy :

(4313 + 4619) (T P + T N ) = × 100 (T P + T N + F P + F N ) (4313 + 4619 + 20 + 2) = 99.46%

Recall : F - Measure :

TP 4313 = × 100 = 99.35% (T P + F N ) (4313 + 28)

(2 × 99.53 × 99.35) (2 × Pr ecision × Recall) = = 99.43 (Pr ecision + Recall) (99.53 + 99.35)

The confusion matrix gives metrics that can also be utilized to calculate the achievement of a project. Accuracy, precision, recall, and F-measure are examples of such measurements. The test dataset (which includes 20% of all articles) is used. The results of the classifier and the selected hyper parameter configurations are reported in Table 2. The results are achieved by utilizing features that characterize the content and structure of each document. Simple, traditional techniques of expressing documents (TF-IDF) do not keep the meanings and relationships of words. When it comes to accuracy, SVM came out with a score of 99.4%. When it comes to precision, SVM came out on top with 99.3%. In terms of recall, SVM came out on top with a score of 99.4%. SVM had the best result for F-Measure, with a score of 99.4%.

5 Conclusion This paper proposes a term frequency extraction for support vector machine-based system for the purpose of detecting fake news, with the goal aimed to determine the optimal techniques along with features for detecting fake news. Thus, it is initiated from researching the area related fake news, it’s effect and ways of detecting it. Then it is devised and a system is built that extracts a set of characteristics that may be used to detect false news from dataset of news that has been pre-processed using feature extraction (TF-IDF algorithms), training the classifier, and using opinion classifier (support vector machine) to classify the fresh data. The findings of this proposed approach are summarized as follows: • Text, author, source, date, and sentiment are greatest features for the detection of fake news. • This procedure provided successfully a noticeable rate.

Term Frequency Tokenization for Fake News Detection

9

• Even though analysis from the text’s sentiment is tempting, it has been more influential in the case regarding opinion mining. • With large datasets and large texts, the TF-IDF performs better compared to other algorithms. • The support vector machine (SVM) seemed to be the better method for detecting fake news, as this provided higher noticeable rate which is allowed for the classification of every piece of information with a degree of confidence.

References 1. Zhang X, Ghorbani AA (2020) An overview of online fake news: characterization, detection, and discussion. Inf Process Manage 57(2) 2. Zhou X, Jain A, Phoha VV, Zafarani R (2020) Fake news early detection: a theory-driven model. Res Pract 1(2):1–25. Article no.: 12 3. Poddar K, Umadevi KS (2019) Comparison of various machine learning models for accurate detection of fake news. In: 2019 Innovations in power and advanced computing technologies (i-PACT), vol 1. IEEE 4. Baarir NF, Djeffal A (2021) Fake News detection Using Machine Learning. In: 2020 2nd International workshop on human-centric smart environments for health and well-being (IHSH), pp 125–130 5. Ozbay FA, Alatas B (2020) Fake news detection within online social media using supervised artificial intelligence algorithms. Phys A Stat Mech Appl 540:123174. ISSN: 0378-4371 6. Ngada O, Haskins B (2020) Fake news detection using content-based features and machine learning. In: 2020 IEEE Asia-Pacific conference on computer science and data engineering (CSDE), pp 1–6 7. Smitha N, Bharath R (2020) Performance comparison of machine learning classifiers for fake news detection. In: 2020 Second international conference on inventive research in computing applications (ICIRCA), pp 696–700 8. Mandical RR, Mamatha N, Shivakumar N, Monica R, Krishna AN (2020) Identification of fake news using machine learning. In: 2020 IEEE international conference on electronics, computing and communication technologies (CONECCT), pp 1–6 9. Reis JCS, Correia A, Murai F, Veloso A, Benevenuto F (2019) Supervised learning for fake news detection. IEEE Intell Syst 34(2):76–8 10. Bharath G, Manikanta KJ, Prakash GB, Sumathi R, Chinnasamy P (2021) Detecting fake news using machine learning algorithms. In: 2021 International conference on computer communication and informatics (ICCCI), pp 1–5 11. Jain MK, Gopalani D, Meena YK, Kumar R (2020) Machine Learning based Fake News Detection using linguistic features and word vector features. In: 2020 IEEE 7th Uttar Pradesh section international conference on electrical, electronics and computer engineering (UPCON), pp 1–6 12. Purevdagva C, Zhao R, Huang PC, Mahoney W (2020) A machine-learning based framework for detection of fake political speech. In: 2020 IEEE 14th international conference on big data science and engineering (Big Data SE), pp 80–87 13. Jiang T, Li JP, Haq AU, Saboor A, Ali A (2021) A Novel stacking approach for accurate detection of fake news. IEEE Access 9:22626–22639 14. Han W, Mehta V (2019) Fake news detection in social networks using machine learning and deep learning: performance evaluation. In: 2019 IEEE international conference on industrial internet (ICII), pp 375–380

10

P. Suresh et al.

15. Manzoor SI, Singla J, Nikita (2019) Fake news detection using machine learning approaches: a systematic review. In: 2019 3rd International conference on trends in electronics and informatics (ICOEI), pp 230–234 16. Zhou X, Zafarani R (2018) Fake news: a survey of research, detection methods, and opportunities. arXiv preprint arXiv:1812.00315 2

Aquaculture Monitoring System Using Internet of Things G. V. R. Kameshwar Rao, T. J. Dhivya Shrilaa, I. Akash, and G. Gugapriya

Abstract Aquaculture comprises an important part of agriculture which has been the pillar in the primary sector of India. Many factors such as temperature, natural calamities and artificial chemicals when monitored irregularly can severely impact the survival of the species. There is more time and energy invested by various scientists across the world to monitor these species in an efficient manner. This is where the idea of an utopian system is initiated to monitor various parameters such as temperature, turbidity, pH and much more through a union of various sensors connected to a microcontroller. The principles of Internet of things and machine learning assist the system to estimate the probability of survival of the species by using the data collected from the sensors. The data which is gathered from the sensors is predicted using the services of IBM and is visualized through a web application that helps the user to interpret data in a coherent manner. Keywords Aquaculture monitoring · Internet of things · Parameter prediction · Machine learning · Chatbot

1 Introduction Aquaculture harvesting was something that was introduced more than 2000 years ago. Fishes and other species are cultivated in small ponds and lakes which contribute to a major share of the agricultural sector. Over the years, India has developed to become a major exporter of shrimps and various other species which has improved the gross domestic product of India [1]. The harvesting of shrimps is very delicate [2] because of the nature of the shrimps not being able to handle the minimal differences with regards to some natural parameters such as temperature, pH, turbidity as well as being unable to withstand the intake of artificial chemicals in minimal quantities. These parameters have been very difficult to monitor daily which becomes a major problem when harvesting these aquatic species [3]. G. V. R. Kameshwar Rao (B) · T. J. Dhivya Shrilaa · I. Akash · G. Gugapriya School of Electronics Engineering, Vellore Institute of Technology, Chennai, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_2

11

12

G. V. R. Kameshwar Rao et al.

The conventional way of monitoring these species was through a doctor who is specialized in the domain of aquatic species. They would have their kit to test out the quality of water as well as the species and can easily find out about their chances of survival when the species is affected by diseases [4]. The flaw with this approach was the monitoring of these parameters daily. Certain species are very vulnerable and can easily die out from the excess of any parameter that is being monitored [5]. The Internet of things has created a massive boost with regards to the approach to solving real-world problems [6]. Users have been able to access their gadgets and monitor various parameters through the medium of the internet. The data which is gathered by various sensors such as a temperature sensor, pH sensor, and much more are transmitted to the IBM Cloud Services through the principles of the internet of things which eases the extraction and visualization of data [7]. With the internet of things, the system that we propose can monitor parameters essential to predict the survival of the species. The parameters that the system is going to monitor apart from temperature and pH would be the values of turbidity and total solvents that are dissolved in the environment. These sensors would record the data that would be beneficial to be used for the process of data telemetry and interpretation by making the best use of the internet of things. With the system being able to interpret the data using the concepts of parameter forecasting, the visualization of the parameters viewed through an application makes the system user-friendly and also assists the user by implementing the services of the chatbot as well along with the application [8]. The culmination of all these objectives would ensure a utopian system that can monitor the parameters required for the survival of aquatic species in an efficient and user-friendly model.

2 Related Works A system has been proposed where in addition to conventional water quality monitoring, research on inter-dependency of the parameters has been considered and is used to reduce the number of sensors involved by rejecting the sensor used for a predictable parameter. In this case, the dissolved oxygen has been neglected as it is expected to be in the normal range if temperature, pH, and conductivity are normal. The results were promising and paved the way for further advancement in soft sensing other parameters. The overall cost of the system is high due to the use of a cell phone to capture the photo of the water to measure the cleanliness and the use of Raspberry Pi for the computation of the photo to detect the color of the water and to store the database [9]. In 2021, the authors [10] discussed other techniques used for soft measurement of water quality parameters using regression and deep learning network. The IPSO (Improved Particle Swarm Optimization)-Least Squares Support Vector Regression (LSSVR) method used to perform a nonlinear prediction model of dissolved oxygen when compared to the traditional Least Squares Support Vector Regression (LSSVR) model, provides better results in terms of reduced root mean square error

Aquaculture Monitoring System Using Internet of Things

13

and mean absolute error values. The Long short-term memory (LSTM) deep learning network performed by the authors shows results up to 98.56% and 98.97% short-term prediction accuracy of pH and water temperature respectively [11]. Android apps have been developed to facilitate the remote monitoring of the water quality parameters where the authors implement a three-layered architecture system consisting of the presentation layer mobile application, data access layer—MySQL database manager, and the devices layer—sensors, actuators, etc. This system uses a general packet radio service module to send the data to the cloud storage and thus is suitable for implementation in remote locations without access to the internet. The mobile application also displays the parameter values in a graph for daily reports. The authors compared the obtained parameter values with the manual tested values and found the variation to be negligible. The overall performance of the system in terms of size, weight, and percentage of survival of the shrimp was observed to be more effective than the manual testing systems [12]. In 2019, the authors have focused on the error percentage of the sensors between laboratory tests and sensor data acquired and forecast of the water quality parameters using an automatic exponential smoothening model using historical values of the parameter thus giving an insight to the farmers to help them in better decision making. As the model used is a time series forecasting method and the parameters are not entirely seasonal with respect to time, the forecast trend obtained is unreliable [13]. Other methods have also been used for the forecast of water quality such as the artificial neural networks (ANNs) for fish farm management. The model forecasts the future data using the historical data and the results obtained are used to perform necessary actions to prevent future catastrophes. The authors construct a wireless sensing network with the use of a general packet radio service combined with virtual private network functionality to collect data on a country-wide scale. The acquired data can be viewed in real-time through the internet [14].

3 Methodology The proposed method consists of 5 subsystems that make up the overall monitoring system. The subsystems are data extraction, data telemetry and visualization, parameter prediction, chatbot, and actuator.

3.1 Data Extraction To assess the quality of the water to maintain a suitable thriving condition for the aquaculture farm, various parameters are required to be monitored. This model aims to measure the following parameters-pH, total dissolved solids (TDS), temperature of water, turbidity and water level. The parameters are chosen based on the dependency

14

G. V. R. Kameshwar Rao et al.

of various hard-to-measure water quality parameters on the crucial and easily measurable parameters. The dissolved oxygen has been neglected since we can assume its normalcy if the pH, temperature, and conductivity are in the normal range. Sensors used to measure these parameters are connected to the microcontroller directly.

3.2 Data Telemetry and Visualization The extracted data is sent over the internet to a cloud platform where it is stored in a database [15]. These values are also sent to an android app with a customer loginbased system. The data is visualized for easy interpretation in the form of graph charts for each parameter and is displayed in the dashboard.

3.3 Parameter Prediction To perform soft sensing of a parameter to effectively neglect the use of its sensor, a machine learning algorithm is used to predict the parameter based on historical data of dependency on other parameters. Various prediction models are used on the training data and applied to testing data [16]. The model which provides the highest accuracy is chosen and is automated to run on the cloud platform using real-time data to predict the parameter which in this case, pH using temperature, turbidity, and water level. The output value is sent to the dashboard in the app.

3.4 Chatbot The user-friendly chatbot service has been provided mainly for solving the queries of the customers. The chatbot has a corresponding response for every action (query) of the user, for example, if the user states that the temperature is low and dissolved oxygen levels are high after heavy rains, then the chatbot suggests that shrimp may be subjected to White Spot Syndrome Virus (WSSV). The chatbot suggests the possible causes of the disease, fluctuation in any vital parameters, resolves doubts regarding the app and the system and helps contact support in case of emergency.

3.5 Water Quality Maintenance Actuators generally act as a response of the system where the data received by sensors determine their response accordingly. In this system the water pump acts as an actuator for maintaining the cleanliness of the water, if the value measured by

Aquaculture Monitoring System Using Internet of Things

15

the turbidity sensor is beyond the normal range, then the water should be replaced immediately with clean water to avoid possible stress to the shrimps. The proposed system performs the automation of this process by replacing the water using pumps which can be controlled by the microcontroller connected via a relay module. The microcontroller is designed to automatically switch on the inlet pump to input clean water and the outlet pump to discharge the unclean water. This process can also be controlled and monitored manually through the app efficiently. The manual controlling has been enabled by the message queuing telemetry transport (MQTT) broker where the app acts as publisher for that particular instance and the NodeMCU being subscriber has subscribed to the topic. According to the payload received the relay is controlled by NodeMCU which in turn controls the pump, thus enabling both manual and automatic controlling of the water pump.

4 Design and Implementation The design of the system (shown in Fig. 1) can be categorized into two parts-the hardware and software. The hardware category consists of sensors that are integrated with the microcontroller and the actuator to respond according to the condition. The software category consists of different IDEs used to program the hardware and develop the machine learning services, chatbot, and android app. A detailed explanation has been discussed in the following sections.

Fig. 1 Block diagram of the system

16

G. V. R. Kameshwar Rao et al.

4.1 Hardware Details Sensors. The Techtonics pH sensor (shown in Fig. 2a) is used to measure the pH level of the water in this work with a measuring range of 0–14 pH with an accuracy of ± 0.1pH. The operating temperature ranges from 0 to 60 °C. The pH electrode is connected via a Bayonet Neil-Concelman (BNC) connector where the signal is amplified by the microcontroller. The suitable pH range for shrimps is 7.9–8.2 pH. It is one of the most important parameters in terms of shrimp monitoring. The DS18B20 temperature sensor (shown in Fig. 2b) is water-resistant with operating temperatures ranging from − 55 to 125 °C and accuracy of ± 0.5 °C. It can be easily controlled via One wire (i.e., single data line required for communication) which makes it one of the most reliable and cheap sensors for usage. The viable range of temperature for shrimps is 27–29 °C which benefits in faster growth and higher reproduction rates. The turbidity sensor (shown in Fig. 2c) measures the value by detecting the amount of light received by the electrode by passing through the water. The operating temperatures range from 5 to 90 °C with a response time of less than 500 ms. The unit for turbidity is NTU which stands for “Nephelometric Turbidity Units”. The optimal turbidity value for shrimp is 21.42 ± 4.43 NTU.

(a) pH sensor

(d) TDS sensor Fig. 2 Sensors and pump

(b) Temperature sensor

(e) Ultrasonic sensor Pump

(c) Turbidity sensor

(f) Water

Aquaculture Monitoring System Using Internet of Things

17

The CentIot total dissolved solvents (TDS) sensor (shown in Fig. 2d) measures the hardness of the water (i.e. salts, minerals, metals, etc.) Turbidity is generally expressed in parts per million (ppm). Its measuring range varies from 0 to 1000 parts per million (ppm) with an accuracy of ± 10% F.S. The optimal total dissolved solvents (TDS) value for shrimp is 200–400 parts per million (ppm). The HC-SR04 Ultrasonic Sensor (shown in Fig. 2e) measures the distance or depth of the object from the sensor using the principles of ultrasonic waves. This sensor can measure up 2–400 cm and can trigger an input pulse for every 10 uS which helps to monitor the species continuously. It can sustain under cold temperatures underwater while utilizing around 20 mA thus being an efficient sensor to monitor these aquatic species. The submersible direct current (DC) water pump (shown in Fig. 2f) pumps the water outward via the plastic probe attached to the pump. The operating range of the pump is 3–6 V and can be controlled by a relay attached to it. The pump has been used in this work to remove the water if the turbidity values surpassed the permissible level of the shrimps. Microcontroller. The NodeMCU development board (WROOM 32) has 38 general purpose input–output (GPIO) pins and consists of an ESP32-S microcontroller (shown in Fig. 3) which supports compatibility with wireless fidelity (Wi-Fi) and low powered Bluetooth (BLE). It has 520 KB of static random-access memory (SRAM), 448 KB of read-only memory (ROM), supports an analog-to-digital converter (ADC), and is programmable via Arduino integrated development environment. This development board has been chosen for the application as it has provided excellent compatibility with wireless fidelity (Wi-Fi) and supports analog-to-digital converter (ADC) conversion. Fig. 3 Node MCU-ESP 32S

18

G. V. R. Kameshwar Rao et al.

4.2 Software Details Arduino IDE. The open-source software is an integrated development environment (IDE) that has been used to program the NodeMCU development board and the code is uploaded via a universal serial board connector. The major advantages of the software are compatible with various versions of development boards, community support, and ease to compile/build the program. Jupyter Notebook. The Jupyter notebook is an integrated development environment (IDE) to develop, train and test the prediction models used in parameter prediction. In this work, we aim to predict the pH of the water using two parametersturbidity and temperature. We perform the comparison of the following multiple regression models-Random Forest regressor, Gradient Boosting regressor, Ridge regressor, Lasso regressor, and Elastic Net regressor. The model that provides the highest R-squared score and least mean average error (MAE) score is chosen as the final model. IBM Cloud Services. IBM cloud, also famously known as Bluemix, is one of the major cloud platforms with various services such as Watson, Internet of Things (Internet of things), database management, analytics, storage, and many more. IBM Cloud is chosen as the cloud service platform for this work to build a chatbot using the Watson Assistant service, to deploy and monitor machine learning models for parameter prediction, and to visualize the parameters in the dashboard. IBM Watson Assistant. It is an advanced processor which performs analytics on the data received and answers the questions accordingly. In this work, we have used it as a user-friendly chatbot to resolve issues with customers. Watson has corresponding action and response to the vast data stored which suggests the best possible solutions for the concerned queries. An information survey is done to gain knowledge on possible information that could be included in the chatbot. The information can be categorized into two divisions-System-related queries and general queries. Systemrelated queries consist of the queries that the users could raise for the successful performance of the system. Example: Hardware system not connected to the internet. General queries consist of the queries that the users could raise related to aquaculture farming. Example: disease detection. This chatbot will be connected to the app through an application program interface (API) so that it can be accessed by the users. Parameter Prediction. The IBM Machine learning service consists of various tools which can be used for deep learning, building and deploying various models, an interactive programming environment, and many more tools which make it user-friendly to build and test various models with vast datasets. The final model is uploaded to the IBM machine learning service through an application program interface to be deployed on the server. The model is then deployed using the real-time data acquired from the sensors and the predicted value of pH is then compared with the acquired sensor value.

Aquaculture Monitoring System Using Internet of Things

19

IBM Cloud Dashboard. The IBM cloud dashboard is generally used to visualize the data which gives great insights about the data and helps in preventing losses. The dashboard is customizable and can be shared with other users too. The data is received in the form of JavaScript object notation (JSON) via the publish-subscribe method. The dashboard displays the live data in the form of a graph and historical data of the parameter values which is sent to the app so that the user can access it from anywhere with internet access. Android Studio. The Android Studio is an integrated development environment (IDE) used for the sole purpose of android application development. The design of the mobile app has been accomplished by coding in Java. The android app consists of a login page in which the users can create a new account or login into their accounts. After logging in, the dashboard page will be displayed which will consist of the realtime sensor data values and pH values predicted using the model. A navigation panel on the top right corner can be used to access other pages-an aquaculture database page used to access the historical data of all the parameters and a water quality maintenance page to monitor or manually control the water inlet and outlet pumps in the system. All the pages will have the chatbot icon in the bottom right corner.

4.3 Overview of Implementation The system (shown in Fig. 4) is powered by a 5 V battery connected to the microcontroller. The sensors are calibrated and connected to the microcontroller which is programmed to acquire the water quality parameter values with the interval of one minute. The microcontroller should be constantly connected to wireless fidelity (Wi-Fi) in order to send the data over to the IBM Watson Internet of things platform which uses the publish-subscribe method. The parameter values are displayed in the IBM dashboard along with the pH value automatically predicted by the machine learning model. This data is sent to the app which also requires constant internet access. The automization of the pump control is done which checks the conditions required to perform the action. The control of the pump can also be done manually through the app using the message queuing telemetry transport (MQTT) protocol. The pump will be run till the conditions are satisfied and the status will be updated in real-time in the app.

4.4 Implementation of pH Prediction Data Collection. Regressor-based machine learning models used for prediction require a dataset consisting of the dependent parameter-pH and the other independent parameters turbidity, temperature, and water level. The datasets required for this purpose are acquired from Kaggle and GitHub repositories [17, 18].

20

G. V. R. Kameshwar Rao et al.

Fig. 4 Block diagram

Data Preparation. The dataset is analyzed for outliers which are removed in Excel sheets. Then the dataset is uploaded to the Jupyter server hosted locally. A Jupyter notebook (shown in Fig. 5) is created on the server and the required modules such as numpy and pandas are imported for visualization. The dataset is imported to a variable and a description of the dataset is obtained for analysis. It can be inferred (shown in Fig. 6) that there are 944 entries in this dataset and the minimum and maximum values of the parameters are in the threshold range. Visualization of the parameters in the dataset is performed using Matplotlib.pyplot and it can be inferred from the output (shown in Fig. 7) that the parameters are correlated to each other. Therefore, a regression model can be used to predict a dependent parameter using the other independent parameter due to the level of correlation between them.

Fig. 5 Required modules are imported

Aquaculture Monitoring System Using Internet of Things

21

Fig. 6 Importing data

Fig. 7 Visualization of the dataset

Choosing and Training the Model. To choose the model that would give the best accuracy for the dataset, five regressor prediction models are trained and tested on the dataset to analyze the accuracy. The models are-Random Forest regressor, Gradient boosting regressor, Ridge regressor, Lasso regressor, and Elastic Net regressor. The dataset is split into training (80% of the dataset) and testing (20% of the dataset) data (shown in Fig. 8). The required modules of the regressors are imported. The regressor models are pipelines (shown in Fig. 9) to automate the machine learning process and the weights are set for each model. The weights are tuned after a number of trial and error and feedback after prediction.

22

G. V. R. Kameshwar Rao et al.

Fig. 8 Splitting the dataset into training and testing data

Fig. 9 Pipeline making and tuning weights

Required modules are imported and the models are fitted (shown in Fig. 10) for the training data. Evaluation and Inference. On performing evaluation and comparing the R-squared score (R2) score and mean absolute error results, it is found that the random forest regressor performs well for the dataset used in this work. The highest R-squared (R2) score and the lowest mean absolute error is obtained in the random forest regressor (RFR) model (shown in Fig. 11). We can infer from the above table that the random forest regressor (RFR) model achieves approximately 97.3% accuracy in predicting the pH for the testing data. Therefore, we choose this as the final model to be used for soft sensing the pH parameter from the real-time inputs in the IBM cloud. Deployment. The model has to be deployed in the IBM cloud in order to request pH prediction and to automize the prediction using the values of independent parameters acquired from the sensor to successfully replace the pH sensor. The deployment is performed using the application program interface (API) client of the IBM Watson Machine Learning service and the required modules are imported. The application

Aquaculture Monitoring System Using Internet of Things

23

Fig. 10 The regressor models are fit for the training data

Fig. 11 Comparison of R2 and mean absolute error scores of pH prediction

program interface key, uniform resource locator (URL) of the server, and the space ID of the service is acquired from the IBM cloud website (shown in Fig. 12). The meta data of the model necessary to deploy in the IBM cloud is defined. The Random Forest model is stored in the ‘BEST_MODEL’ variable. The ‘MODEL_NAME’ and ‘DEPLOYMENT_NAME’ are provided. The software specification UID is acquired through a function that searches the ID using the python version of the Jupyter notebook which is version 3.8. The model is then stored in the IBM cloud repository along with the training dataset (shown in Fig. 13).

24

G. V. R. Kameshwar Rao et al.

Fig. 12 Setting the credentials and space ID

Fig. 13 Storing the model in IBM cloud repository

5 Results and Discussion A system that is capable of extracting data values from various sensors that are imbued with the microcontroller along with the services of IBM would provide the results with which the user can cultivate the species in a productive manner. The data is visualized in the application which makes the user to interpret the values easily.

Aquaculture Monitoring System Using Internet of Things

25

Fig. 14 Data visualization in IBM Watson

5.1 Sensor Data Visualization in IBM Watson Internet of Things Platform The data which is recorded in their individual sensors is transferred to the IBM Watson Platform by accessing the microcontroller through the medium of Internet. In this IBM Watson Platform (shown in Fig. 14), the data that is transferred from the microcontroller is visualized in their individual units such as temperature being recorded in the units of Celsius and turbidity being recorded in the units of nephelometric turbidity units (NTU). This helps to understand the influence of the parameters and their impact on the aquatic species.

5.2 Parameter Prediction On evaluating the Random Forest model for pH prediction using the testing data, 0.973 R2 score and 0.0397 mean absolute error score is achieved. From the above graph and tabulation (shown in Figs. 15a, b), it can be seen that the difference between the original and predicted values of the pH are negligible and therefore we can safely implement the system without using the pH sensor by soft sensing the parameter using temperature, turbidity, and water level parameters. This model is uploaded to the IBM cloud using the application program interface (API) client and deployed (shown in Fig. 16). After uploading the model in IBM, the test tab can be used to test the model by giving the input fields in JavaScript object notation (JSON) format and the results return the predicted value of the pH. This value is displayed in the IBM dashboard and also sent to the android app.

26

G. V. R. Kameshwar Rao et al.

Fig. 15 a Original pH values versus predicted pH values, b examples of original pH values versus predicted pH values

Fig. 16 Deployment in IBM cloud

Aquaculture Monitoring System Using Internet of Things

27

Fig. 17 Chatbot interface

5.3 Chatbot The Chatbot service is essential for the user to understand the nuances about the operations of the system. A user might encounter some trouble with regards to the system either in the form of hardware malfunction, data visualization, connectivity issues, or more problems. For resolving these issues as well as to learn about a particular component of a function of the system, the chatbot would be beneficial in terms of providing the solution (shown in Fig. 17). It is built in with standard commands with regards to the configuration of the system as well as to provide information with regards to the main components of the system. The chatbot follows a user-friendly approach which makes the user interact with the chatbot service with ease. It also helps in connecting the user with the customer support team when the issue is beyond the scope of standard queries.

5.4 Android App The login page (shown in Fig. 18a) is the page where the users sign up or login into their accounts. Upon successful login, the app displays the dashboard (shown in Fig. 18b) with the real-time updates of all the parameters monitored in their system. The chatbot icon is displayed in the bottom right corner of all the pages. The navigation panel allows the users to navigate between the database and the maintenance pages.

28

G. V. R. Kameshwar Rao et al.

a) Login screen

b) Dashboard

Fig. 18 Android app

6 Conclusion and Future Scope This work designs and implements a monitoring system that not only monitors and displays the results to the farmers but also is a step towards efficient soft sensing through its prediction model, therefore, reducing the costs involved in the hardware and its maintenance. The system also provides customer support and information regarding the important aspects of aquaculture thereby enabling the farmers to make well-informed decisions. The app lets the farmer access the state of their aquaculture farm from anywhere with the use of the internet and solves the dependence on manual testing. Aquaculture monitoring, though a well-researched field, provides ample opportunities for research and new inventions. More actuators can be included in the system and automated thereby fully leveraging the latest technologies. The scope of research in the area of soft sensing is immense and can be improvised to predict all the important parameters.

Aquaculture Monitoring System Using Internet of Things

29

References 1. Fisheries statistics. https://dof.gov.in/sites/default/files/2021-08/Handbook_on_Fishery_2020. pdf 2. Nisar U, Zhang H, Navghan M et al. (2021) Comparative analysis of profitability and resource use efficiency between Penaeus monodon and Litopenaeus vannamei in India. PLOSONE J 3. Dixon W, Chiswell B (1996) Review of aquatic monitoring program design. Pergamon J 30(9) 4. Nash CE (2011) The history of aquaculture 5. Quach A, Murray F, Morrison-Saunders A (2017) The vulnerability of shrimp farming income to climate change events. Int J Clim Change Strat Manage. Ca Mau, Vietnam 6. Kumar S, Tiwari P, Zymbler M (2019) Internet of Things is a revolutionary approach for future technology enhancement. J Big Data 7. Kochut A, Deng Y, Head M, Munson J, Sailer A et.al. (2011) Evolution of the IBM Cloud: enabling an enterprise cloud services ecosystem. IBM J Res Dev 8. Tratar LF, Mojškerc B, Toman A (2016) Demand forecasting with four-parameter exponential smoothing. Int J Prod Econ Part A 181 9. Saha S, Hasan Rajib R, Kabir S (2018) IoT based automated fish farm aquaculture monitoring system. In: International conference on innovations in science, engineering and technology (ICISET) 10. Wang C, Li Z, Wang T, Xu X et al. (2021) Intelligent fish farm—the future of aquaculture. Aquac Int 11. Hu Z, Zhang Y, Zhao Y et al. (2019) A water quality prediction method based on the deep LSTM network considering correlation in smart mariculture. Sens J 12. Gómez-Chabla R, Real-Avilés K, Delgado-Vera C, Chávez C, Vera-Lucio N (2018) Monitoring system for shrimp farming. In: CITI 2018. Communications in computer and information science, vol 883. Springer, Cham 13. Cordova-Rozas M, Aucapuri-Lecarnaque J, Shiguihara-Juárez P (2019) a cloud monitoring system for aquaculture using IoT. In: 2019 IEEE sciences and humanities international research conference (SHIRCON) 14. Li D, Liu S (2013) Remote monitoring of water quality for intensive fish culture. In: Mukhopadhyay S, Mason A (eds) Smart sensors for real-time water quality monitoring. Smart sensors, measurement and instrumentation, vol 4. Springer, Berlin, Heidelberg 15. Christina G (Sept2021) A review on microstrip patch antenna performance improvement techniques on various applications. J Trends Comput Sci Smart Technol 03(03):175–189 16. Hamdan YB (2021) Construction of statistical SVM based recognition model for handwritten character recognition. J Inf Technol 3(02):92–107 17. Dataset for assessing the various parameters of aquaculture monitoring. https://github.com/ pkErbynn/IoT-WQMS 18. Dataset for aquaculture monitoring. https://www.kaggle.com/alexisaucapuri/fishfarm-monito ring

A Comprehensive Study and Implementation of Memory Malware Analysis with Its Application for the Case Study of CRIDEX Digvijay Singh and Rajesh Yadav

Abstract The advancement in technology and the rising demand for the internet have witnessed an increased cyber threat which can be highlighted as a major challenge. Cyber criminals make use of malware programmes to fulfil the malicious purpose. Malware is software coded and designed to bring harm to the target machine by various means. This paper focuses on describing memory analysis techniques under malware analysis with a comparative analysis of different tools. The present paper highlights the application to the case study of CRIDEX malware for a better understanding through a practical implementation of appropriate tools. With its initial appearance as CRIDEX, the bank stealing trojan has evolved in the past decade and has been witnessed to spread malware infection through its new variants as discovered in the past year for which it has been selected as an appropriate candidate to be analysed for understanding its basic working. Keywords Malware · Memory analysis · Artifacts · Memory dump · Cridex

1 Introduction Malware is an intrusive software purposefully designed by cybercriminals with intentions of a potential threat. Malwares can steal, corrupt and destroy data. Each malware has a special characteristic which makes it distinct from another malware. Memory analysis is a powerful technique of Malware analysis, it deals with capturing physical memory of the infected state of a machine and later performing a complete analysis of that memory. This technique has been proven efficient for performing malware analysis and is also broadly categorized under memory forensics [1]. It came into existence in recent years. D. Singh · R. Yadav (B) Department of Computer Science, BML Munjal University, Gurgaon, Haryana, India e-mail: [email protected] D. Singh e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_3

31

32

D. Singh and R. Yadav

Memory analysis fulfils the purpose of performing a digital investigation thus giving useful artifacts. An analyst uses this technique by analysing a dumped memory of an infected state of the target machine. Once the malware is executed, it makes modifications to the system and these are obtained via artifacts similar to the ones being used for investigation [2]. This research describes an approach which can be applied by an analyst to perform memory analysis on different malware samples. We will be showing a comparison of different tools to differentiate the features and identify the appropriate tools capable of giving faster and accurate results. We have included Cridex malware to analyse the infected state of the memory dump captured post its execution. The remainder of this paper is organised as follows: Sect. 2 talks about the background work, Sect. 3 states the Literature survey. In Sect. 4 we will discuss the Methodology and Sect. 5 will show the implementation of the case study followed by a comparative analysis of the different tools. Section 6 highlights the results obtained by the analysis and Sect. 7 concludes the paper.

2 Background Work Memory analysis technique is popular for providing a comprehensive view of the infected system. Being efficient and delivering accurate results, it is encouraged by both analysts and forensic investigators. It is divided in two phases—Memory Acquisition and Memory Dump Analysis. Memory acquisition refers to the process of acquiring or capturing a physical memory state of an infected system by using appropriate tools [3] followed by the second phase of analysing that memory dump using suitable features provided by the tool. Memory Analysis Memory is an essential component used for storing information. The Central Processing Unit works synchronously with the memory unit for the storage operations. When we talk about Malware Memory Analysis, we focus on the RAM [4]. It is responsible for allowing the data to read or write independent of the physical memory. For memory management, we typically have a stack and heap memory. Stack works on the principle of push and pop operation, making it grow eventually. For Dynamic memory allocation, a heap memory is used. • Memory Acquisition—This stage deals with the dumping of an infected memory state. Once a malware is executed and infects the system, it gets loaded in the RAM. It is essential to dump this memory using a suitable tool for completing further analysis. • Memory Dump Analysis—This stage deals with analyzing a dumped memory state. An analyst uses the acquired state for performing operations to extract information like process information, network related information and much more [5]. For understanding this stage any memory dump can be used. There are various

A Comprehensive Study and Implementation of Memory Malware …

33

Table 1 Different types of malware and their impacts Types of malware

Impact

Virus Trojan Spyware Bot Ransomware Wiper Worm Rootkit

Infects application by exploiting vulnerabilities Performs malicious activities in the background Monitors user activities while aiming for sensitive information Automates infection spreading tactics Demands a ransom against any sort of attack Usually wipes off the data and memory Spread the infection over a network and flood the network servers Usually conceals the presence and is accessible remotely

databases containing memory dump of an infected state caused by malware. Once the memory state is acquired the tools are used for gathering as much information to analyze the impact caused by the malware. Different types of malware and their impacts are shown in Table 1. Cridex Malware To strengthen our application-based study we have included a Cridex malware sample in our case study. It is banking malware capable of stealing the banking-credentials and other vital information [6]. It is categorized as a Trojan which can gain complete access to the infected system. It has a capability of spreading via copying itself in different drives.

3 Related Work The research papers based on Malware Analysis focuses broadly on different techniques of analyzing malware. Most common techniques involve Static and Dynamic analysis; however, some researchers have highlighted the importance of Hybrid and Memory analysis. Seo et al. [7] has described different ways of gathering memory data. The authors have included dumping techniques and highlighted the difficulties while dumping a pure memory and also discussed the challenges of dumping a physical memory of RAM more than 4 GB. Manson et al. [8] has shown a comparison between three different tools- Autopsy, FTK imager and Encase. The comparative analysis by the authors mentions that the image import was done fastest by Encase but the user experience was reported not smooth for the analysts. Apart from Encase, the other two tools were much easier to be used and FTK imager is best suited for windows. The authors have highlighted its intuitive interface. Okolica et al. [9] has introduced the CMAT tool. It is capable of parsing memory dump to identify all processes. It works on the principle of extracting data from a live memory by compiling the responses into a suitable form for further analysis. Carvajal et al. [10] have done a comparison of six tools which are used for analyzing the memory. The parameters for

34

D. Singh and R. Yadav

comparison are based on the interface experience, artifacts, processing time, reports. The study tells that the GUI based tools left more artifacts than the CLI based. The authors of [11] have done a survey on tools and techniques for memory acquisition and analysis which specifically deal with windows-based systems. The authors of [12] have illustrated the prevalent concepts of creating memory snapshot and has also shown methods of analyzing the memory data. Kim et al. [13] have developed a techniue which can be used for classification of malwares. The importance of malware classification is highlighted in order to bring attention for identification of malware and their families. It helps to understand different patterns related to a malware infection. The author of [14] has proposed an approach to identify malicious android applications by providing a complete mechanism. Current techniques were compared using different datasets. The algorithm can be considered as a complete in terms of solving real challenges linked to android categorization. The authors of [15] have introduced memory aware cloud scheduling for a faster scheduling. The idea is to improve overall security and maintaining an error free process. It was tested through simulations with total value of 50 applications. Ori et al. [16] have shown a survey to overview different dynamic analysis methods. The authors have also focused on classification and behavioural analysis. The paper discuses about Memory forensic technique which brings attention towards this approach of analysing the malware.

4 Memory Malware Analysis Approach Memory Analysis is about capturing a malware infected state of memory by acquiring RAW memory and further analyzing the memory dump to perform an assessment of different parameters. An analyst retrieves vital information and associates it with the artifact which could be left unnoticed if not analyzed. Memory analysis is performed mainly for three reasons • An analyst can retrieve real time data from the physical memory. • The encrypted data is decrypted once it is acquired in the RAW memory. • Information of higher significance can be obtained by this technique, which can be difficult while performing static and dynamic analysis [17]. Memory Acquisition Acquisition is the technique of acquiring/Dumping a physical state of memory. When a system is infected by malware, there are equal chances of the system going to a dead or alive state. The execution of malware makes the necessary changes as coded by the threat actor. If the system is alive then different tools capable of capturing the memory can be used for making a memory dump and if the system is dead then a forensic disc controller can be used to collect the data and the remaining process remains the same. To ensure that there are no changes in the Memory Dump, a hash value can be calculated and verified as required [18]. Some of the common tools

A Comprehensive Study and Implementation of Memory Malware …

35

which can be used to acquire the physical memory state are- FTK imager, Autopsy Forensic Browser, Encase, Nuix, Hibernation Recon and Forensic explorer. Memory Dump analysis Once an analyst acquires a memory dump the next step is to analyze it. Memory Dump analysis is the technique of analyzing an infected memory dump to extract vital information that will be used to analyze a malware. This technique focusses on analysing the artefacts from the memory dump [19] as in Fig. 1. Depending on the size of the memory dump, appropriate tools analyze the memory file and the processing time is directly proportional to the size of that particular dump. It is also observed that a larger dump is capable of producing more artifacts than a smaller one. This phase can be applied on any available memory dump. An analyst can examine memory dump either by acquiring through capturing or simply use an existing memory dump available on the internet for the desired malware sample. Potential information which can be extracted through this technique generally includes: suspicious processes, DDL and handles, network related information, injected code, Dump processes and drivers, kernel memory etc. Some of the common tools which can be used to analyze a memory dump are- Volatility, Autopsy Forensic Browser, Caine, Rekall, Cado Live and Redline.

Fig. 1 Memory malware analysis process

36

D. Singh and R. Yadav

5 Available Tools FTK imager: FTK Imager is one of the best open-source tools used for creation of disk images. It provides an interactive user interface with advanced capabilities. An analyst uses this tool to acquire images without making changes in the system. The tool is capable of recovering the files which are deleted from the Recycle Bin but the changes are not overwritten on the drive. It allows the user to mount an image to view the contents of the acquired image as on the original drive. Autopsy Forensic Browser: Autopsy Forensic Browser is a GUI based application which is capable of analysing hard drives and smartphones. It offers file type identification, registry analysis, data extraction from Android and is used for military investigations. It is extensively used for mobile device and digital media forensics. EnCase: EnCase tool is considered as a Global standard in digital investigation. It efficiently conducts the needs of investigators and analysts by using a repeatable and defensible process. It supports a broad range of file systems. It offers automation in the analysis of artifacts which proves to be time saving unlike other memory analysis tools. The wide range of features offered by the tool are not completely free. Forensic Explorer: Forensic Explorer (FEX) is a tool which is mainly used for preservation and analysis of memory. It gives a suitable experience to its users including law enforcement, government and military agencies. It offers a simple and interactive GUI experience providing multiple functionalities at a time. The tool is capable of processing a larger volume of data and provides detailed report covering different aspects like- File and Registry Analysis, Boot Virtualization etc. Volatility: Volatility is an open-source framework for memory analysis. Currently it is one of the most popular tools supporting Windows, Mac and Linux. The pythonbased framework is best suited to analyse the RAM in 32/64-bit systems. It allows us to analyse different parameters including- date and time of the image captured, Network related information, Memory address, Process related information, libraries loaded and kernel modules. Volatility is recommended for memory analysis because of its faster and efficient algorithm of analysing RAM dumps. Since it is compatible for different operating systems, it has a larger repository storing profiles for different Operating Systems. Memoryze: Memoryze is a memory analysis tool. It was formerly known as MANDIANT Free Agent. This tool is not only limited towards acquisition of physical memory rather it can perform advanced memory analysis while the system is in a live state. It can image a complete address of a process to a disk including dlls and exe’s. Also, it can verify the digital signature of dlls and exe’s.

A Comprehensive Study and Implementation of Memory Malware …

37

6 CRIDEX Case Study Memory Acquisition. This step deals with capturing memory of the system. The idea is to make a memory dump of the malware infected. FTK imager was used to acquire RAW. memory. User Interface of FTK imager tool is shown in Fig. 2. Memory Analysis This step deals with analysis of the memory dump as acquired previously. The idea is to gain information by monitoring memory changes. It reveals process related information which helps an analyst to examine suspicious activities and produce maximum artifacts. We have used Volatility framework to fulfil the purpose of analysing memory. A view of command line interface is shown in Fig. 3. To begin with the analysis, we provided the cridex.mem file and applied various operations supported by volatility framework. Firstly, we used “volatility.exe -f cridex.mem imageinfo” to determine the profile based on kDBG search. By using this command, we identified that the tool hints on targeted OS as WindowsXP as shown in Fig. 4. The next step was to list an overview of the running processes using “volatility.exe -f cridex.mem pslist”. More than 15 running processes were identified as shown in Fig. 5. Out of the 17 processes listed in the above image 13 of them are executed by Windows OS, however 4 processes reader_s1.exe, alg.exe and wuaclt.exe (listed twice) are not part of windows services hence suspicious.

Fig. 2 User Interface of FTK imager tool

38

D. Singh and R. Yadav

Fig. 3 User ınterface of volatility framework

Fig. 4 Profiles associated with the acquired memory dump

Further to listing running processes we displayed the the processes as a parent and child tree using “volatility.exe -f cridex.mem pstree”. Similar to pslist this command also highlights the suspicious processes but alg.exe and wuauclt.exe are indicated as a part of winlogon.exe which makes them non suspicious however reader_s1.exe is shown as standalone as shown in Fig. 6. Process reader_s1.exe has Pid 1640 and PPid 1484 which means it is directly running under explorer.exe. The focus was towards identifying this suspicious process. We used “volatility.exe -f cridex.mem connscan” to get the network related information and we were able to identify the local and remote address linked to Pid 1484 as in Fig. 7. By using “volatility.exe -f cridex.mem sockets” we identified that the port number of PID:1484 is 1038 which provides the service for Message tracking Query Protocol as shown below in Fig. 8.

A Comprehensive Study and Implementation of Memory Malware …

Fig. 5 Process list generated by volatility framework

Fig. 6 Prcoess tree indicating parent child relation between processes

Fig. 7 Network information fetched through the memory

39

40

D. Singh and R. Yadav

Fig. 8 Scoket information displaying port information for respective PID’s

Since Pid 1640 was identified to use PPid 1484 which was marked suspicious in the aforementioned steps, then we used “volatility.exe -f cridex.mem dlllist -p 1640” to get a list of DLLs. The information extracted through this command shows that the process uses kernel 32 and network dll as shown below in Fig. 9. With the strengthening of our doubt against process with Pid 1640 we used “volatility.exe -f cridex.mem procdump -p 1640 –dum-dir.” to dump the file in the existing directory as shown below im Fig. 10. As soon as the command was executed, executable.1640.exe file appeared on desktop as shown in Fig. 11. To analyse the nature of the suspicious file we uploaded the sample on virus total engine and identified that the file was a trojan as shown in Fig. 12.

Fig. 9 List of DLLs obtained through volatility

A Comprehensive Study and Implementation of Memory Malware …

41

Fig. 10 Creation of a dump file for process 1640

Fig. 11 An exe file being generated on desktop after dumping the process

Fig. 12 API based detection using virus total

The previous step confirms our doubt of the suspicious process with Pid:1640. Now we used memdump to create a.dmp file to further extract relevant information as shown in Fig. 13. Once the “.dmp” file was generated, we used strings command line utility to extract the strings from the file to identify more information. The trojan consisted of several strings containing website addresses of different banks. Similar bank related

42

D. Singh and R. Yadav

Fig. 13 Creation of.dmp file

information was revealed through other strings. A view of extracted strings have been shown in Fig. 14. The experimental findings indicate that the process with PID 1484 is suspicious, the process uses port 1038. Port number 1038 provides a Message tracking Query protocol. On dumping the file application, an exe file was visible on desktop. Executable 1640.exe was used as an input on virus total. The tool Classified the file as a trojan which was further examined by doing string analysis. On performing string analysis, it was revealed that the trojan was focussed towards stealing banking credentials and other confidential information. The strings consisted a big list of bank websites. Fig. 14 Banking information present in the form of strings

A Comprehensive Study and Implementation of Memory Malware …

43

7 Conclusion Memory Analysis technique is an emerging technique widely used in Malware Analysis. It helps to acquire and analyse an infected memory dump. Memory Analysis and Forensic are similar techniques dealing with analysis of a memory file. Our work has described the memory analysis technique by highlighting its importance in Malware Analysis. We have identified and surveyed different tools capable of performing memory acquisition and memory dump analysis. Memory Malware Analysis follows a 2-step approach- fetching a RAW memory, analysing the memory dump. Our work has included an analysis of different tools capable of performing Memory Malware Analysis. For a better understanding, we have included CRIDEX malware as a part of our case study. Our experimental analysis was focussed towards analysing the memory dump of Cridex infected state and our experimental findings tells that the malware is a trojan. We identified that Cridex was responsible for execution of a bank stealing trojan. The file executable-1640.exe was categorised as a trojan on Virus Total engine and the.dmp file was used to extract strings denoting several banking websites and other legal information related to bank accounts.

References 1. Rathnayaka C, Jamdagni A (2017) An efficient approach for advanced malware analysis using memory forensic technique. [online] IEEE Xplore. Available at: https://ieeexplore.ieee.org/abs tract/document/8029568/. [Accessed 19 Jun 2020] 2. Sihwail R, Omar K, Zainol Ariffin K, Al Afghani S (2019) Malware detection approach based on artifacts in memory image and dynamic analysis. Appl Sci 9(18):3680 3. Stüttgen J, Cohen M (2013) Anti-forensic resilient memory acquisition. Digit Investig 10:S105– S115 4. Ravindra Sali V, Khanuja HK (2018) RAM forensics: the analysis and extraction of malicious processes from memory ımage using GUI based memory forensic toolkit. In: 2018 Fourth ınternational conference on computing communication control and automation (ICCUBEA) 5. Guangqi L, Lianhai W, Shuhui Z, Shujiang X, Lei Z (2014). Memory dump and forensic analysis based on virtual machine. [online] IEEE Xplore. Available at: https://ieeexplore.ieee. org/document/6885969?arnumber=6885969. [Accessed 25 Apr 2022]. 6. Webopedia (2015) What is cridex? [online] Available at: https://www.webopedia.com/defini tions/cridex-malware. [Accessed 25 Apr. 2022]. 7. Seo J, Lee S, Shon T (2013) A study on memory dump analysis based on digital forensic tools. Peer-to-Peer Netw Appl 8(4):694–703 8. Manson D, Carlin A, Ramos S, Gyger A, Kaufman M, Treichelt J (2007) Is the open way a better way? Digital forensics using open source tools. In: 2007 40th Annual Hawaii ınternational conference on system sciences (HICSS’07). 9. Okolica J, Peterson G (2010) A compiled memory analysis tool. Adv. Digit Forensics VI:195– 204 10. Carvajal L, Varol C, Lei Chen (2013) Tools for collecting volatile data: a survey study. [online] IEEE Xplore. Available at: https://ieeexplore.ieee.org/abstract/document/6557293. [Accessed 16 May 2020].

44

D. Singh and R. Yadav

11. Dolan-Gavitt B (2008) Forensic analysis of the Windows registry in memory. Digit Investig 5:S26–S32 12. Vömel S, Freiling FC (2011) A survey of main memory acquisition and analysis techniques for the windows operating system. Digit Investig 8(1):3–22 13. Kim M, Kim D, Hwang C, Cho S, Han S, Park M (2021) Machine-learning-based android malware family classification using built-in and custom permissions. Appl Sci 11:10244. https:// doi.org/10.3390/app112110244 14. Vivekanandam B (2021) Design an adaptive hybrid approach for genetic algorithm to detect effective malware detection in android division. J Ubiquitous Comput Commun Technol 3(2):135–149. https://doi.org/10.36548/jucct.2021.2.006 15. Haoxiang W, Smys S (2020) Secure and optimized cloud-based cyber-physical systems with memory-aware scheduling scheme. J Trends Comput Sci Smart Technol 2:141–147. https:// doi.org/10.36548/jtcsst.2020.3.003 16. Or-Meir O, Nissim N, Elovici Y, Rokach L (2019) Dynamic malware analysis in the modern era—a state of the art survey. ACM Comput Surv 52(5):48. Article 88 (Sept 2020). https://doi. org/10.1145/3329786 17. Armstrong B (n.d.) About dump encryption [online] docs.microsoft.com. Available at: https:// docs.microsoft.com/en-us/windows-server/virtualization/hyper-v/manage/about-dump-enc ryption [Accessed 25 Apr 2022] 18. Faiz MN, Prabowo WA (2018) Comparison of Acquisition Software for Digital Forensics Purposes. Kinetik: Game Technol Inf Syst Comput Netw Comput Electron Control [online] 37– 44. Available at: http://repository.ittelkom-pwt.ac.id/5808/1/WAP%20Paper%20KINETIK% 20Comparison%20of%20Acquisition%20Software.pdf [Accessed 25 Apr 2021] 19. Dayalamurthy D (2013) Forensic memory dump analysis and recovery of the artefacts of using tor bundle browser—the need. [online] Available at: https://core.ac.uk/download/pdf/ 41535535.pdf [Accessed 25 Apr 2022]

IoT Based Anti Poaching of Trees and Protection of Forest E. V. Kameswararao, M. Jaya Shankar, T. V. Sai Lokesh, and E. Terence

Abstract Theft of the world’s most valuable trees, such as sandalwood, lumber, teak, rosewood, and pinewood, presents a huge danger to forest resources, causes substantial economic harm, and has a terrible impact on the ecosystem across the globe. These trees are very pricey and scarce around the globe. These trees are employed in both medicinal and cosmetics research. To stop such smuggling and conserve the world’s forests, various preventative measures must be created. Many incidences of tree cutting occur as a result of the higher amount of money involved in selling such trees. This study presents an anti-poaching system based on IoT and WSN technologies. Three sensors are used in the structural framework: a tilt detector (to detect the tendency of a tree while it is being cut), a fire sensor and a smoke detector (to detect timberland fires), and a sound detector Detection of even the sound of a tree being hacked down may be used to catch illegal loggers, for example WSN technology is commonly employed in remote monitoring applications (where monitoring is difficult). Keywords Threat · Devasting · Restrict · Preventive measures · Frame work · Unlawful logging

1 Introduction Timber theft of ecologically and commercially significant tree species in wooded regions—such as Teakwood, coast redwood, Sandalwood, Pine, and Rosewood— has expanded considerably as a result of poaching or smuggling. To address these issues, several players, including the Indian government, have launched a number of initiatives. Anti-poaching observers and/or private/government security guards may E. V. Kameswararao (B) · M. Jaya Shankar · T. V. Sai Lokesh · E. Terence Department of ECE, Hindustan Institute of Technology and Science, Chennai, India e-mail: [email protected] E. Terence e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_4

45

46

E. V. Kameswararao et al.

be recruited, trained, and deployed around woods to help with this. Strict punishment for convicted criminals, as well as unique incentives for smuggling operations (Five Year Plan 2012–2017), were all aimed at putting a stop to the threat. However, the punitive measures have mostly been ineffectual, and poachers have thrived as a result [1–4]. The most effective option is to Build a real-time, wireless sensor network (WSN) and data recording system, a complex and low-cost contemporary technology that will make monitoring more robust, effective, and feasible. The Wireless Sensor Network (WSN) is a novel technology. That is already being utilised in a variety of industrial applications, including monitoring, disaster observation (such as forest fires), living space surveillance, maintenance, security and control, and remote monitoring applications. WSNs are commonly utilised in forest regions for detecting forest fires, detecting tree smuggling, environmental monitoring, and other purposes [5]. Wireless Sensor Networks are simple to set up and maintain, and they save money by eliminating the usage of expensive wires. We can adopt the method that was used to reduce the degree of smuggling in the forest regions with the aid of WSN and other detectors. Poaching isn’t only a problem in India; it’s also a problem in China, Australia, and certain African countries. In India, sandalwood, rosewood, and pine food tree cost between 12,000 and 13,000 INR per kg, whereas Red Sanders costs INR 10 focus for each ton in the global market. The Indian sandalwood tree has become more interesting as of late, provoking the Indian government to endeavor to restrict sandalwood exports in order to combat the tree’s potential hardship. According to the government, the maximum allowable weight gain for an individual is three.8 kg. If the tree is presently under government control, its removal will be banned, regardless of whether it is for personal or sanctuary reasons. The primary purpose of this project is to create a framework that may be used to prevent sandalwood, pine, and teak timber trees from being stolen. The proposed work is arranged in the following order: i. ii. iii. iv. v. vi.

Introduction Related works Proposed work Results Conclusion and Future work References.

2 Related Works Khandare et al presents the design of a system for detection of vibration and sound for prevention of cutting of trees, detection of temperature and fire for prevention of forest fires. The sensing device can sense the vibration, sound and temperature, fire and then information sent them over zig-bee network to forest office. The fundamental disadvantage of this system is its reliance on batteries for continuous monitoring and data transmission to a central server [6].

IoT Based Anti Poaching of Trees and Protection of Forest

47

Baraddi a technique that may be utilised to prevent smuggling. Three sensors are used in the design system: a sound sensor, a tilt sensor, and a temperature sensor. The Blynk App continually monitors the data collected by these sensors. In terms of sensors, the relay switch turns on the sensors’ output devices. A signal is sounded for the slant and sound sensors, and a water siphon is incited for the temperature sensor. The Wi-Fi module saves the created information in Blynk Server. The main drawback of this system is it Can’t be communicated to long distance [7]. Narasimman this system is a reliable and low-cost wireless vibration monitoring system. Vibration was measured using a 3-axis digital output MEMS Accelerometer sensor. Vibrations as low as 0.0156 g and as high as 8 g may be detected by the Accelerometer detector, with 1 g equalling 9.81 m/s2. The AT-mega328p microcontroller on the Arduino UNO board is used to connect with the MEMS Accelerometer detector. The primary disadvantage of this technology is that it is susceptible to higher frequency noise [8]. Rohan devised a system that can be used to prevent tree smuggling, thereby forestalling deforestation and keeping up with natural strength. Each tree is furnished with an electronic division, which includes sensors such as the Raspberry Pi, accelerometer sensor, Micro Controller, Fire sensor, Flex Sensor, and GSM/GPRS module. The flex sensor and accelerometer sensor will be used to detect forest tree cutting. GPRS/GSM modules will be used to communicate between the trees and the server. The biggest disadvantage of this method is the price. In comparison to other systems, it is somewhat costly [9]. Mhaske developed a method that can be used to control and restrict tree smuggling, which would eliminate deforestation, reduce wood theft, and maintain environmental stability, all of which would assist with lightening one of the difficulties of a worldwide temperature alteration. A Micro Controller, Flex Detector, Accelerometer Sensor, Temperature Detector, Zigbee and GSM module are just some of the sensors and controllers that are included into each tree. A flex sensor and an accelerometer sensor will be used to detect tree chopping in order to avert it. GSM modules will communicate with the server and the trees. TWO phases comprise the system: 1. The tree unit The main server unit is referred to as B. The primary disadvantage of this technology is that it is susceptible to higher frequency noise [10]. Raghavendra et al the fundamental goal of this method is to reduce and prevent smuggling, conserve precious trees, and minimise wood theft in request to keep a decent eco-framework by diminishing deforestation. The technique use GPS technology to pinpoint the position of the tree where poaching occurs. The system uses a chip (micro controller board) with embedded sensors (flex sensor and fire sensor) that are controlled via IoT. These sensors monitor and control parameters such as tree tilting, burning, and cutting, and they can be accessed via an Android App on an Android phone. The biggest disadvantage of this technique is that it’s tough to keep track of the forest [11]. Arunprasath et al author created a methodology that may be used to reduce sneaking. Detecting illegal logging, for example, involves listening for noises generated when cutting down a tree. Other uses for the temperature and tilt sensors include detecting forest fires (to recognise the tendency of tree when its being cut). The web

48

E. V. Kameswararao et al.

page/app continuously monitors the data generated by these sensors. Sensors’ yield devices are activated by a hand-off switch, which is a relay switch. A ringer is established for the inclination and sound sensors, while a water siphon is impelled for the temperature sensor. The Wi-Fi module is utilized to save the information made in the cloud server. The fundamental disadvantage of this system is its reliance on batteries for continuous monitoring and data transmission to the central server [12]. Sudharani et al to limit sneaking and screen trees, author structure can be made using gyro locater (to perceive the inclination of tree when it is being cut), temperature pointer (NTC 10 k thermistor), Wi-Fi Module (esp8266), and GSM Module. The data accumulated by these sensors is ceaselessly noticed using the ThingSpeak cloud stage, which is sent from a microcontroller (Arduino Uno). The data is saved in ThingSpeak Server through the Wi-Fi module. The main disadvantage of this system is that it is inflexible. Long-distance communication is not possible [13]. Hingane et al author designed a system using some sensors like flex sensor, pH Sensor, Fire Sensor, Ultrasonic Sensor, Node MCU(microprocessor) Relay, Water pump, Buzzer to monitoring the forest continuously for avoid the smuggling and save the forest from forest fire when forest is in fire. The main drawback of this system is fetched (cost). It is very expensive as compared to other systems [14]. Kirangond suggested an anti-poaching system that includes an archetype of an IoT model that monitors trees and alerts the base station in [15]. When the tilt values change, the base station and the registered cell phone number get the coordinates of the poached tree. In comparison to the current plans, the suggested model reduces human involvement. GSM module, Raspberry Pi board, and accelerometer sensor GPS module are included in the proposed model, which will aid in tree monitoring. This document presents the proposed model’s working theory, which will aid in the improvement of the current system while also acting as a resiliency for the proposed system. Small places, such as private property, and vast areas, such as national reserves, may both benefit from the suggested approach. The system’s biggest flaw is that it’s far away. When compared to other systems, it is rather pricey [16].

3 Proposed Work The proposed system will be made up of two modules: one with sensors and controllers that will be installed in the tree, and the other with an Android phone or computer. We created an application that would receive sensor data on a constant basis. This is an IoT-based project in which sensor data is regularly uploaded to the biodots cloud. Tilt sensor and sound sensor is used to determine whether the tree is cut down or not. Similarly, fire/flame sensor and smoke sensor is used to determine whether the forest is on fire or not. The sensed data continuously uploaded in the biodots cloud for every 30 s. The block diagram of the proposed architecture is shown in Fig. 1.

IoT Based Anti Poaching of Trees and Protection of Forest

49

Battery LCD Tilt sensor

Flame sensor

Cloud (Ubidots)

Arduino Nano Atmega 328p Node MCU

Sound sensor

Fig. 1 Block diagram

The major objective of this project is to construct a portable wireless sensor node that may be utilised in a Wireless Sensor Network. An Android phone or computer with sensors and controls will be embedded in a tree as part of a planned two-modular system. We created an application that would receive sensor data on a constant basis. This is an IoT project in which we regularly upload sensor data to the ubidots cloud. Tilt sensor and sound sensor is used to determine whether the tree is cut down or not similarly Fire/Flame sensor and smoke sensor is used to determine whether the forest is on fire or not. The sensed data continuously uploaded in the ubidots cloud for every 30 s. We can able to know the actual GPS location of the sensors. When a sound is detected, the sound sensor generates an output voltage signal. A microcontroller receives voltage and begins doing the required processing. The process flow diagram is shown in Fig. 2. Noise levels in decibels (dBs) may be measured by the sound sensor at frequencies between 3 and 6 kHz, which is about where the human ear has sensitivity. It reacts to sound intensity in the same manner that the human ear does. It accurately monitors sound levels across a single range of 55–110 dB to within 3 dB. There are noise levels in the forest owing to trees and animals, so if a tree falls, the noise level is abnormal, which is reported to the forest officer, and there is a distinction between tree fall down due to natural disasters and tree fall down due to unlawful methods. Table 1 shows the threshold value of tilt and smoke sensor and Table 2 on flame and sound sensor. A smoke sensorr is a device that senses smoke (air particles of different gases or fire), typically as an indicator of fire. Smoke sensor do not have a listed spacing. They have a recommended spacing of 30 feet between sensors. However, smoke sensor

50

E. V. Kameswararao et al.

Start

sensor inputs

No

No

No

Change in tilt

If High fire

No

Smoke detects

Sound detects

Yes

Yes

Yes Yes

Obtain the tree status (Controller) Node MCU Update data in ubidots for every 30secs

Stop Fig. 2 Flow chart

Table 1 Threshold values of tilt and smoke sensor

Sensor name

Threshold value

Tilt sensor

430(When it is NOT tilted)

Smoke sensor

212

Table 2 Threshold values of flame and sound sensor Sensor name

High

Low

Flame sensor

1 (Presence of fire)

0

Sound sensor

1 (Presence of sound)

0

can be installed up to 41 feet. A fire locator is a sensor that recognizes and answers the presence of a fire or fire, making fire identification conceivable. They operate at moderate speed with a range of up to 200–250 feet from the flame source.

IoT Based Anti Poaching of Trees and Protection of Forest

51

Fig. 3 Node MCU architecture

Initially, when it is NOT tilted position, it shows the fixed threshold value 430 in the LCD as well as in the web pageWhen it is inclined in a particular direction or angle, it shows value of tilting position of a tilt sensor on the LCD as well as in the web page. If we tilt the sensor in the left side direction the values are decreased. Similarly, if we tilt that sensor in right side direction the values are increased. It can detect the values of tilting position of a tree in between 20 and 60°. Node MCU architecture is shown in Fig. 3. The sensed data, sends to controller. After receiving the data, the controller starts processing and perform the necessary operations in it. The Wi-Fi module receives data as well. The controller sends data to the Wi-Fi module, which subsequently sends it to the cloud through the internet (Ubidots). We can monitor the data in Ubidots cloud.

4 Results Figures 4 and 5 shows the only either Left—Right side direction or Front—Back side direction. The tilt sensor works from 0 to −70° as well as from 0 to 70°. The remaining angle position −70 to 180 and 70–180° is uncertain region. If the tilt sensor is NOT tilted position (i.e., at 0°), it shows the actual threshold value 430 in the LCD and as well as in the Web page. If the sensor tilted left (i.e., it bends to −20–−70°), it shows the value of decreasing threshold and varies according to bending angle and if it is tilted right (i.e., 20–70°), it shows the value of increasing threshold and varies according to bending value. It is shown in the Web page. If the angle between −20 and −70 and 20–70°, the tree is tilted heavily and we can conclude that it indicates like there is something illegal happening. Similarly in front and back directions also same. The front direction indicates the right-side direction

52

E. V. Kameswararao et al.

and backside direction indicates the left side direction. The results are getting from right and front directions are same, similarly in left and backside directions are same. Table 3 shows the practical values of tilt sensor.

Fig. 4 Result

Fig. 5 Working position of tilt sensor

IoT Based Anti Poaching of Trees and Protection of Forest

53

Table 3 Data table of tilt sensor Direction

Actual angle

Practical value

Left (or) Back Right (or) Front

0° (When it is NOT tilted) −45° −60° 0° (When it is NOT tilted) 45° 60°

430 362 315 430 488 510

Table 4 Data table of sound sensor Sensor Name

Practical values

Condition

Sound sensor

0 1 1 0

Low High High Low

If sound sensor detects sound, it shows high value ‘1’.If there is no sound, it shows the value ‘0’on the Web page. Table 4 shows the practical values of sound sensor. If flame sensor detects any fire near by its, it shows the ‘High’ value ‘1’. If there is no fire it shows ‘Low’ value ‘0’.in the LCD as well as in web page. Table 5 shows the practical values of flame sensor. The smoke sensor threshold value is 212. If the smoke sensor value shows below threshold value, there is no much amount of smoke and less impact of fire. If the value is reaches above threshold value, there is a high chances of forest fire. Table 6 shows the output values of smoke sensor. Figure 6 shows the LCD output. Table 5 Data table of flame sensor Sensor name

Practical values

Condition

Flame sensor

1 1 1 1

High High High High

Table 6 data table of smoke sensor Sensor name

Threshold value

Practical value

Smoke sensor

212

216 228 237 236

54

E. V. Kameswararao et al.

Fig. 6 LCD (output display)

5 Conclusion and Future Works 5.1 Conclusion A portable wireless sensor network has been developed for save the forest from forest fire, deforestation and some important trees like Sandalwood, Teakwood, Pine and Rosewood etc. The motive of this design for keeping trespassers away and save the forest from timber mafia. For this reason, Different sensors like Tilt, Sound, Smoke, Fire sensors and Arduino Nano board, LCD, Wi-Fi module is used. To avoid the Trees cutting, Tilt and Sound sensor is used. To save the forest from forest fire, fire/flame and Smoke sensor is used. For remote terminal through wireless media, Wi-Fi module is used. We can monitor the data on LCD as well as in the Ubidots cloud.

5.2 Future Works This study may be improved in the future by using a multi-node network with microphones, motion detector sensors, data collection systems may be improved by using temperature and human or animal interference sensors as well as sensors to monitor fires.

IoT Based Anti Poaching of Trees and Protection of Forest

55

References 1. Sudha BS (2018) Forest monitoring system using wireless sensor network. Int J Adv Sci Res Eng 4(4), E-ISSN: 2454–8006 2. Dalvi A (2018) Undetected detective to protect the forest trees against poaching using WSN technology, vol 3, no 6 3. Gaikwad S (2015) Design WSN Node for Protection of Forest Trees Against Poaching based on Zigbee. In: 2015 IEEE international conference on electronics, computing and communication technologies (CONECCT) 4. Ghousia Sultana B (2018) IOT based anti-poaching alarm system for trees in forest using wireless sensor network. Int J Adv Res Comput Sci 9(3) 5. Hamza HC (2013) In: Tree theft control system 2013 texas instruments India educators conference. 6. Khandare A, Malve M, Kulkarni A (2014) Zigbee based wireless sensor network (WSN) for protection high cost trees in from fire and poaching. IRJET 7. Baraddi P, Hanchinal N, Jadhav R, Shushma, Banni R (2020) IOT Based Anti-Poaching Alarm System for trees in forest. IJERT. https://www.ijitee.org/wp-content/uploads/papers/v8i6s/F60 510486S19.pdf 8. Narasimman V, Anand, Anandha kumar, Krishnan T (2018) Design of a wsn node for forest trees against poaching. IJARSE. http://www.ijarse.com/images/fullpdf/1521184541_DACE2062ij arse.pdf 9. Rohan (2018) Real time forest anti-smuggling monitoring system based on IOT using GPRS and GSM. Int J Res Eng Appl Manage (IJREAM) (ICSGUPSTM 2018). ISSN: 2454-9150 10. Mhaske D (2016) Anti-smuggling system for trees in forest using flex sensor with GSM & Zigbee network. Int J Adv Res Comput Commun Eng 5(4) 11. Raghavendra (2019) IoT based anti-smuggling system for trees in forest. Int J Sci Res Rev 8(6) 12. Arunprasath, Naveenraj MT, Srinivasan R, Jeevabarathi C (2019) IoT based anti-poaching alarm system for trees in forest. (IJITEE) 8(6S), ISSN: 2278-3075 13. Sudharani CH, Shilpa (2019) IoT based anti-poaching of trees. Int J Eng Adv Technol (IJEAT) 8(5), ISSN: 2249-8958 14. M.C.Hingane, Snehal Choudhari, Vandana Datta Ingale, Sonali Awachare, “Anti-poaching Alarm System For Tree in Forest”, IJEDR 2019 | Volume 7, Issue 4 | ISSN: 2321–9939. 15. Kirangond, Mudanuri Chaithanya K, Shellagi PR, Sampath Kumar NG, Nivedha S (2020) Anti poaching of trees using Raspberry Pi. IJESC 10(6), ISSN 2321-3361 © 2020 16. Narhari R, Kotkar ME (2014) (ESD AND VLSI) Anti-smuggling system for trees in forest using flex sensor and Zigbee. Int J Adv Res Comput Eng Technol (IJARCET) 3(9)

Artificial Intelligence Based Efficient Activity Recognition with Real Time Implementation for ATM Security S. Srinivasan, AL. Vallikannu, Adapa Sankar Ganesh, Iragamreddy Raj Kumar, and Beereddy Venu Gopal

Abstract Recognizing human activities plays a substantial role in human-to-human and human-to-computer interactions. Recognizing human activities from video sequences or pictures is a difficult task because of troubles, such as history clutter, partial occlusion, modifications in scale, viewpoint, lights and look. Human action is difficult to classify as a time series. Predicting a person’s movements is a part of this. In this paper, the KTH video dataset is used for designing the system. Feature extraction methods like optical flow and spatiotemporal techniques are being utilized to extract the features. Triple stacked autoencoders are used for clusterization to reduce the data dimensions. An efficient BoW vector feature extraction method is used for extracting text data, by which data is obtained for training the model. A deep learning algorithm such as VGG19 is used to determine and classify the activities of a human. The objective of this efficient model is to apply as an ATM surveillance as a camera module fixed in the room to perform constant surveillance. The Police department can have an mobile application through which they can monitor and desist any unwanted human activities happening in the ATM. Keywords Human activity recognition · ATM security · VGG19 deep learning · Sequence processing · BoW vector feature extraction · KTH video dataset · ATM app S. Srinivasan (B) · AL. Vallikannu Department of ECE, Hindustan Institute of Technology and Science, Chennai, Tamil Nadu, India e-mail: [email protected] AL. Vallikannu e-mail: [email protected] A. S. Ganesh · I. R. Kumar · B. V. Gopal Hindustan Institute of Technology and Science, Chennai, Tamil Nadu, India e-mail: [email protected] I. R. Kumar e-mail: [email protected] B. V. Gopal e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_5

57

58

S. Srinivasan et al.

1 Introduction Human Activity Recognition is a broad subject of study concerned with identifying the precise motion or movement of a person based on sensor information. Activities are regularly done standard movements which include walking, standing, sleeping, and sitting. They may also be extra targeted on other activities such as the varieties of activities executed in a kitchen or in a workplace. The sensor data can be remotely recorded, inclusive via video, radar, or other wireless methods. Alternatively, information can be recorded straightforwardly regarding the matter, for example, by carrying custom hardware or smartphones that have accelerometers and gyroscopes. Fitness and health tracking technologies such as cell phones are now widely available and affordable. As a result, the cost of acquiring data from these devices is lower, not unusual, and consequently is a more generally studied version of the general activity recognition hassle. It is the goal of human activity recognition software to analyse video or still images to determine what people were doing at a given time. Human activity recognition systems are impacted by this fact and work to accurately categorise input data into the appropriate activity category. To classify human activities, terms like “gestures,” “atomic motions,” “human-to-object,” “group movements,” “behaviours,” and “events” are used, all of which are based on the degree of complexity they involve. When it comes to gestures, they are considered to be primordial bodily components that may be linked to a certain movement of a person. It is possible to describe an activity in terms of its “atoms,” which are the parts that make up the whole. Human-tohuman or human-to-item interactions include people interacting with other people or things. Group activities are those that are carried out by a group of people. A person’s emotional, personality, and psychological condition may be communicated via their body movements. The ultimate definition of events is that they are excessively staged actions that represent social movements among persons and imply the aim or social standing of those involved.

2 Literature Review A Double Stacked Autoencoder with implanted grouping (DSAFEC) and a BOW building strategy in view of the DSAFEC to limit computational intricacy and kill determination limitations (B-DSAFEC) has been offered in [1]. The DSAFEC used video feature points to generate BOWs for human activity identification by predicting the probability of feature point cluster assignment. Soft clustering instead of hard clustering gives each feature point to a number of clusters with the greatest probability, rather than assigning each feature point to only one cluster. Trial discoveries on three benchmark human action datasets showed that the B-DSAFEC outflanks

Artificial Intelligence Based Efficient Activity Recognition with Real …

59

five different procedures created utilizing either ordinary bunching techniques or profound grouping strategies. In [2], PrivHome, a privacy-preserving solution, was proposed. For smart home systems, it offers authentication, safe data storage, and inquiry. For security reasons, PrivHome ensures information classification as well as element and information confirmation for all gatherings associated with the information transmission. Additionally, the gateway offers privacy-preserving queries so that neither the service provider nor the gateway can access the data. The idea of protection safeguarding requests for savvy home frameworks has never been raised, essentially not as far as anyone is concerned. Entity and key sharing protocols as well as a searchable encryption mechanism have been added to the system, were included in the method. Because both protocols rely entirely on symmetric cryptographic techniques, the scheme was practicable. There are comparisons to current smart home security methods based on data collected via experiments and modelling, to demonstrate the efficiency and effectiveness of the method. Using FMCW radar, a novel Dynamic Range-Doppler Trajectory (DRDT) approach was proposed in [3], for recognising continuous human movements in real time in a variety of situations simulating real-life conditions. Continuous motions can be separated and processed as single events using this method. The backscattered signals were first used to create range-Doppler frames, in which each map is composed of a progression of reach Doppler maps. For constant following of human developments in time and reach and Doppler areas, the DRDT was constructed from these frames. The DRDT map was then used to find and segregate each individual human motion using a peak search approach. Final results were obtained by extracting and integrating range, Doppler, Radar Cross-Section (RCS), dispersion characteristics into a machine learning classifier using a multidomain fusion technique. Even when the distance, angle of view, and direction of the view, as well as the diversity of individuals, were all altered, this method achieved accurate and robust recognition. Extensive testing has been carried out to demonstrate its practicality and superiority. Continuous categorization yielded an average accuracy of 91.9%. In order to avoid identification of source cameras based on PRNU, the research [4] suggested a number of solutions. Forced seam carving is an example of a new counter forensics approach. By eliminating seams to the point that most uncarved picture blocks are less than 50 × 50 pixels, the method eliminateed the need for PRNUbased identification of the source camera. Even if the image’s uncarved blocks are smaller than the 50 × 50 pixel minimum, the results of this research reveal, source attribution can still be done given numerous seams carved photographs from the same camera. The community broker function was employed in study [5] to integrate community services, providing electronic information services, reducing the burden of community managers, and enhancing connections between the community and its surrounding environment. Integrating various sensors and gadgets in the house to provide energy management, scenario information, and security are all parts of a

60

S. Srinivasan et al.

house intranet that were developed using a decent touch board and a home regulatory framework. Local area PCs, as well as connectivity to other communities and home networks, are part of the end-to-end community solution (e.g., video cameras and building automation equipment). With the usage of standard interface devices, numerous in-home displays may be achieved by separating the logic and user interfaces. A Light Field Camera as a fresh perspective for detecting face show assaults was described in [6]. A light field camera’s interesting ability to make different profundity (or concentration) pictures in a solitary catch is due to its capability to record both the direction and intensity of each incoming ray. A new method for detecting presentation assaults was therefore described, which examined focus fluctuations across multiple LFC-generated depth (or focus) pictures. Using a light field camera, first a fresh face artefact database of 80 individuals was gathered. Facial artefacts were made by mimicking two typical assaults: picture printing and electronic screen attacks. Tests on the light field artefact database have shown that the suggested PAD technique outperforms many well-known state-of-the art strategies. A recurrent neural network approach is used to analyse the dicriminative training of data analytics was proposed in [7] which performed well in activitity recognition. A most compact and efficient model with feature descriptors and classifiers proposed by Cozar et al. [8] produced a comparatively good results on activity recognition. A deep learning clustering metric learning method was suggested by Law et al. [9] and Dizaji et al. [10]. They obtained the gradient loss function which involves less complexity for training a model. The method learnt future representations and embeddings required for clusters. Detecting human action from video datasets are very complex and the same was addressed in [11] providing spatio-temporal localization. Activities of learners in an online class is monitored by a deep learning method proposed by Vivekanandam [12] in which the emotional behaviors of human is recognised with satisfied computational speed. A Neural Network based suspicious human activity detection system was proposed in [13] where they used temporal and qualitative knowledge in detection of anomaly of a human [14]. To summarize, the activity recognition of human has been discussed and analysed for various reasons like security and detection of any anomaly etc. in many literature using Neural Network and Machine Learning algorithm. Many of the works are lagging in identifying the recognition of human activity interms of accuracy. Also many works used dual stacked auto encoders which for video dataset is insufficient to extract the data.

3 Proposed Model In this paper, the human activity recognition is determined by which the user’s activity in an ATM can be effectively identified by which an abnormal action can be determined. So, the first stage is to gather data from different sources, and then divide this data into preparing and testing datasets, where the preparation dataset is kept discrete

Artificial Intelligence Based Efficient Activity Recognition with Real …

61

and the testing dataset is used to prepare the model. After then, the dataset is supplemented with new data to make it larger. When these datasets have been analysed, they are aligned using various procedures. In order to create the system, the KTH video dataset is employed. After preprocessing the dataset, architectural training can be performed. Feature extraction methods like optical flow and spatiotemporal techniques are being utilized to extract the features. Triple layered auto encoders are used to minimise the size of the data. Using efficient BoW vector feature extraction, data for training the model are gathered from text data. An algorithm like VGG19 is then used to identify and categorise a person’s activities. As a camera module is permanently installed in the room, this effective model can be used in ATM monitoring. In the event that the ATM camera detects an abnormality in the human’s activities, an alarm is raised to the police. When the ATM camera detects anomalous behaviours of a person attempting to break the ATM’s security, an alarm is delivered and live streaming is enabled via a mobile application made using react native. Proposed System Architecture is shown in Fig. 1. In this research work, VGG19 is used for classifying the activities from obtaining the features as it shows accuracy close to 92%. VGG19 model has faster training speed, fewer training samples per time, and higher accuracy. VGG is a deep Convolutional Neural Network used to classify the images. VGG19 is a variation of the VGG model that has 19 layers in total (16 convolution layers, 3 Fully connected layer, 5 MaxPool layers and 1 SoftMax layer). VGG11, VGG16, and other variations are all forms of VGG. VGG19 has a total of 19.6 billion FLOPs. To train this network, an RGB picture with a fixed size of (224 * 224) is fed, and therefore the matrix is shaped as (224, 224, 3). For preprocessing, simply the mean RGB worth of every

Fig. 1 Proposed system architecture

62

S. Srinivasan et al.

Fig. 2 Algorithm to classify the activity of a human in an ATM

pixel, determined across the entire preparation set, is disposed off. (3 × 3) measured bits with a step size of 1 pixel is utilized, permitting them to traverse the entire picture idea. The picture’s spatial goal is safeguarded by the utilization of spatial cushioning. Sride 2 is utilized to create greatest pooling on a 2 by 2 pixel window. In order to increase classification accuracy and speed up computations, the model is further refined by using the Rectified Linear Unit (ReLu), a non-linear function that introduced non-linearity into the model. To classify 1000-way ILSVRCs using 1000 channels, a softmax function is used in the last layer of three fully linked layers, the first two of which are 4096 bytes long. The algorithm for identifying the human activity in this work is given in Fig. 2. React native is being utilised to create a mobile app for this work. React Native is a framework for creating JavaScript code from a hierarchy of UI components. Using this framework, a native-looking mobile app can be created for both iOS and Android. The mobile industry has grown at a rate that has never been seen before. By the year 2020, mobile apps are expected to produce income of over 188 billion dollars in the United States via app stores, advertising, and in-app purchases. A highquality app is needed for both personal and corporate usage, with various displays, simple navigation, and a decent design. On the other hand, developing high-quality, high-performing native applications takes a long time compared to developing crossplatform apps, which is quicker but sacrifices quality and support. If you’re looking to develop a high-quality app in a short period of time, React Native seems to be a suitable alternative. The development environment for mobile apps in this paper is Android Studio. It depends on JetBrain’s IntelliJ IDEA programming and is custom-made solely for Android improvement. Android Studio is the authority IDE for Google’s Android working framework. In 2020, it will be a free download or a membership based

Artificial Intelligence Based Efficient Activity Recognition with Real …

63

assistance for clients of Windows, macOS, and Linux working frameworks. As the major IDE for native Android application development, it replaces Eclipse Android Development Tools (E-ADT).

4 Results and Dıscussıon The dataset collection involves the process of collecting normal human activity that include walking, posing etc. which a human performs inside an ATM room as shown in Fig. 3i, ii. Apart from the regular activity listed above, other activities like running, jogging and hand waving are also collected for training dataset and performed several times by 25 subjects in four different scenarios: outdoors, outdoors with scale variation, outdoors with different clothes and indoors. The database contains 2391 sequences. All sequences were taken over homogeneous backgrounds with a static camera with 25fps frame rate. The sequences are downsampled to the spatial resolution of 160 × 120 pixels and have a length of four seconds in average. The validation matrics the human activity is shown in Table. 1. Then these datasets are pre-processed by converting the images into required size format so that it can be made ready for training with the model. Training is performed using CNN algorithm. After training the stacked autoencoder, an accuracy of approximately 99% is achieved. This means that the stacked autoencoders can recreate the original input signal with about 99% accuracy. The decrease in loss during the training and the increse in accuracy during training are depicted in Figs. 4 and 5 respectively. From the above results, it can be seen that the triple stack autoencoder algorithm has been successfully implemented to automatically recognize the human activity. In the existing system, the system uses dual stacked auto encoders which for video dataset is insufficient to extract the data and the system does not provide an effective technological solution for real time deployment. In order to overcome this problem, a camera module that is permanently installed in the ATM room provides a costeffective solution for ATM security. It’s able to identify a normal human activity as shown in Fig. 6 inside an ATM room, and in the event that the ATM video detects odd behaviour from a person attempting to get into the machine as shown in Fig. 7, an alarm is forwarded to authorities. Hence, the purpose of this project is to develop and deploy an action recognition and gesture recognition system that can detect human activities automatically. Figure 6 explains the application of such human normal activity in an ATM room (drawing money by standing in front of the machine) to identify what is happening inside and it helps to improve the security. In Fig. 7, when a person tries to perform an unwanted activity like break open the ATM machine etc., it is identified by the algorithm, the video is captured and security alert is sent to nearby Police station to safeguard the ATM. Through live streming, a Police staff can view in the ATM app, which is shown in Fig. 8.

64

S. Srinivasan et al.

Fig. 3 i Dataset collected (1), ii dataset collected (2)

5 Conclusion The Human Activity Recognition model has been created and successfully implemented. It recognizes the human activity in the ATM room and sends the notification to the police if any abnormal activity is detected. The police are provided with an application called ATM app, in which the police is able to watch the live streaming of what is happening inside the ATM. Spatio temporal and optical flow features are

Artificial Intelligence Based Efficient Activity Recognition with Real …

65

Table 1 Validation matrics Activity

Precision

Recall

F1-score

Support

Walking

1.00

1.00

1.00

70

Boxing

1.00

1.00

1.00

37

Hand clapping

1.00

1.00

1.00

41

Hand waving

1.00

1.00

1.00

55

Standing

1.00

0.99

0.99

207

Breaking

0.93

0.96

0.95

27

0.99

437

Accuracy Macro avg

0.99

0.99

0.99

437

Weighted avg

0.99

0.99

0.99

437

Fig. 4 Decrease in loss during training

extracted. Triple stacked auto encoders are used for clusterization. Deep Learning VGG19 algorithm is used for training the model. Thus the recognition is done. Results are examined and classified as normal and abnormal activity, which may be used in an ATM for real time surveillance.

66

Fig. 5 Increase in accuracy during training

Fig. 6 Normal activity

S. Srinivasan et al.

Artificial Intelligence Based Efficient Activity Recognition with Real …

Fig. 7 Abnormal activity

Fig. 8 Live streaming of ATM app

67

68

S. Srinivasan et al.

References 1. Wang T, Ng WWY, Li J, Wu Q, Zhang S, Nugent C, Shewell C (2021) A deep clustering via automatic feature embedded learning for human activity recognition. J Latex Class Files. IEEE 2. Poh GS, Gope P, Ning J (2019) PrivHome: privacy-preserving authenticated communication in smart home environment. IEEE Trans Dependable Secure Comput 18(3):1095–1107 3. Ding C, Hong H, Zou Y, Chu H, Zhu X, Fioranelli F, Le Kernec J, Li C (2019) Continuous human motion recognition with a dynamic range-doppler trajectory method based on FMCW radar. IEEE Trans Geosci Remote Sens 57(9):6821–6831 4. Taspinar S, Mohanty M, Memon N (2017) PRNU-based camera attribution from multiple seam-carved images. IEEE Trans Inf Forensics and Secur 12(12):3065–3080 5. Lee YT, Hsiao WH, Huang CM, Seng-Cho TC (2016) An integrated cloud-based smart home management system with community hierarchy. IEEE Trans Consum Electron 62 6. Raghavendra R, Raja KB, Busch C (2015) Presentation attack detection for face recognition using light field camera. IEEE Trans Image Process. Norwegian Biometric Laboratory, Gjøvik University College, Norway 7. Richard A, Gall J (2017) A bag-of-words equivalent recurrent neural network for action recognition. Comput Vis Image Underst 156:79–91 8. Cozar JR, González-Linares JM, Guil N, Hernández R, Heredia Y (2012) Visual words selection for human action classification. In: 2012 International conference on high performance computing simulation (HPCS), pp 188–194 9. Law MT, Urtasun R, Zemel RS (2017) Deep spectral clustering learning. In: Proceedings of the 34th international conference on machine learning, vol 70, pp 1985–1994 10. Dizaji KG, Herandi A, Deng C, Cai W, Huang H (2017) Deep clustering via joint convolutional autoencoder embedding and relative entropy minimization. In: IEEE international conference on computer vision, pp 5747–5756 11. Shao L, Jones S, Li X (2014) Efficient search and localization of human actions in video databases. IEEE Trans Circuits Syst Video Technol 24(3):504–512 12. Vivekanandam B (2020) Evaluation of activity monitoring algorithm based on smart approaches. J Electron 2(03):175–181 13. Bhambri P, Bagga S, Priya D, Singh H, Singh H, Dhiman HK (2020) Suspicious human activity detection system. J IoT Soc Mob Anal Cloud 2(4) 14. Herath S, Harandi M, Porikli F (2016) Going deeper into action recognition: survey. Image Vis Comput 60:4–21

Terror Attack Classification with the Application of Orange Data Mining Tool and Neo4j Sandbox Ankit Raj, Suchitra A. Khoje, and Sagar Bhilaji Shinde

Abstract There is no universally accepted definition of terrorism. Terrorism and its ramifications have every once in a while caused massive death and destruction around the world. Current cutting-edge technologies, such as machine learning and deep learning, can predict and classify such attacks efficiently. The major difficulties observed in implementing these strategies are a lack of consistent and clean data, as well as programming knowledge in Python and R. Inconsistent data can be resolved by incorporating graph database features into the dataset, and Python programming can be replaced with the orange data mining tool. As a part of data processing and manipulation software, orange data mining tool employs a machine learning model in a non-coding context. This research study has attempted to replicate the results by using the orange tool and Neo4j Sandbox. In this study, a non-coding approach was used to classify terror attacks by using the orange data mining tool, and the use of graph embeddings as dataset features have assisted in eliminating the problems associated with inconsistent data. The dataset was then subjected to machine learning techniques such as Random Forest, Decision Tree, Support Vector Machine, Naive Bayes, Gradient Boosting, KNN, and Adaboost to classify the terror attacks. Random Forest and Gradient Boosting are the models that can achieve an accuracy score, recall, precision, and F1 score greater than 90%. Keywords Neo4j Sandbox · Orange data mining tool · GTD · Machine learning · Graphs

A. Raj (B) · S. A. Khoje MIT–World Peace University, Pune, Maharashtra, India e-mail: [email protected] S. A. Khoje e-mail: [email protected] S. B. Shinde NMVPM’s Nutan College of Engineering and Research, Pune, Maharashtra, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_6

69

70

A. Raj et al.

1 Introduction Relational and non-relational databases are the two main types of databases. A nonrelational form of database is known as graph database. A graph database or graph is a higher dimensional data representation where nodes and relationships are used instead of rows and columns [1]. While the characteristics of those nodes are the rows of the relational database, which indicate the number of entries in a dataset and the nodes in a graph are the entities that represent a column or attribute of the relational database [2]. The global terrorism database managed by the University of Maryland is the dataset used in the proposed research study. Working in the field of machine learning requires a strong working knowledge of the python programming language. Without any python programming experience, Orange Tool provides the flexibility to work in the domain of machine learning [3]. Here, GTD is used to develop a graph database for the proposed project. This graph database has millions of relationships between its thousands of nodes. The graph data science library found in the Neo4j Sandbox plugin was used to calculate certain properties of the graph database, including degree, centrality, and node embedding. Seven machine learning models, including decision tree, random forest, gradient boosting, KNN, SVM, Naive Bayes, and Adaboost were applied to the dataset. According to different performance metrics like AUC Score, accuracy, recall, and F1 score, the best model will be selected. The prediction results were displayed by using the confusion matrix in the orange tool [3].

1.1 Dataset The University of Maryland-owned global terrorism database remain as the source for research data. The dataset is a compilation of every act of terrorism that has taken place around the globe between the year 1970 and 2019 in a relational dataset format. The attributes in the dataset include the timing and location of the assault, the type of weapon used, target type, causation, and more. There are 136 attributes in the dataset with two lakh entries of the terror incidents [4]. We sorted and filtered the data set due to computing resource constraints and selected 10 instances per year from 1970 to 2020. Consequently, there were 500 records in the sample dataset. Figure 1 shows the geographical spots on the world map, where terror attacks have occurred in the past. It can be seen that the South Asia region is the most terror attack-prone region on the globe [5]. The event ˙ID, event location, event time, event date, event day, event month, event year, longitude, latitude, specificity, proximity, attack type, target type, gun type, weapon type, and others are the attributes of the GTD dataset [5].

Terror Attack Classification with the Application of Orange Data …

71

Fig. 1 GTD map for showing terror attacks worldwide. Source www.start.umd.edu

1.2 Graph and Neo4j Neo4j is a JavaScript-based tool for creating and manipulating graphs. CQL or cipher query language is used for its operation [6]. The graph data science library in Neo4j may be used to apply various algorithms to the graph. Neo4j’s computed embedding may be exported as a CSV file. The machine learning model will use the estimated embedding, degree, and centrality as significant features [6].

2 Literature Survey Neo4j is a javascript-based tool used for creating and manipulating graph databases that use the Cypher query language, or CQL. The manipulation and mathematical operations on the graph database are made simpler by the preloaded plugins, such as the graph data science library (GDS) and awesome procedures on Cipher (APC). According to Felix Melchor Santos Lopez and Eulogio Guillermo Santos De La Cruz Neo4j gives the database atomicity, consistency, isolation, and durability (ACID), hence it is an excellent substitute for traditional SQL (Relational Database) [1]. For machine learning applications, Orange is considered as a data mining tool that serves as a substitute for the Python and R programming languages. With many

72

A. Raj et al.

machine learning algorithms, including supervised learning techniques and unsupervised learning approaches, Orange tool was most recently released in 2016 by including a huge library for data preprocessing along with the utilization of data imputation block for removing null values from the dataset, PCA (Principle component analysis) block is commonly used for performing dimension reduction, wherein the data preprocessor block is used for data scaling, and support all the preprocessing algorithms. To show the machine learning use of Orange data mining tool, Musa Peker, Osman Ozkaraka, and Ali Sasar implemented five machine learning models on a diabetic dataset obtained from Dalaman State Hospital of Turkey [2]. Today’s market offers a variety of data mining technologies, including R programming language, Rapid Miner, WEKA, Orange, and Kinme. Rapid Miner is language agnostic, whereas orange was created using C, C++ , Cython, and Python. Orange offers more freedom to the developers by offering them a load model block so that they may create their own models and send them to the orange tool [3]. The benefits and downsides of various data mining technologies were thoroughly compared by Ranjan et al. [3]. The end-user can implement a variety of machine learning models provided by WEKA using the java programming language. With a data training percentage of 66% and test data percentage of 34%, Ghada M. Tolan attempted to use machine learning models by including Naive Bayes, K-nearest neighbour, C4.5, ID3, and support vector machine. The dataset majorly used for terror attack classification is the global terror attack database, which is a copyright of the University of Maryland. It consists of 136 attributes and two lakh entries of incidents from the year 1970 to 2015. WEKA has both machine learning and deep learning support and is an open-source platform issued under GNU general public license [7]. Two of the most popular algorithms for classifying terror attacks are the decision tree and random forest algorithms. While the decision tree has never demonstrated accuracy above 75%, random forest algorithm with modified hyper-parameters has consistently demonstrated results above 90% [8]. Although GDBMS are now more widely accepted by data analysts, they nevertheless have their own drawbacks, such as high computer power requirements, longer calculation times, and more complicated algorithms when dataset sizes grow. Due to the large number of libraries that are filled with graph data science and their connection with the python programming language, Neo4j, Orient DB, and Titan are considered as the most promising graph database management technologies [9]. Neo4j operates twenty times more efficiently than conventional RDBMS, such as Postgre, when compared to the two types of RDBMS. Both the relational database and the graph database have their own advantages and thus it is impossible to say which is quicker because it relies entirely on the application for which it is being used [6]. More than twenty graph database solutions are now available on the market, including Orient DB, Arango DB White DB, Graph DB, Azure Cosmos DB, Fauna DB, Tiger Graph, Neo4j, Velocity DB, Memgraph, Titan, and many others. Of these, Neo4j and Tiger Graph are the two that perform the best. These graph databases are frequently used in the field of biomedical engineering to record patient names, identification numbers, diagnosis, and treatment information.

Terror Attack Classification with the Application of Orange Data …

73

The advantages of graph databases over relational databases have also been demonstrated with a thorough comparison of the graph database frameworks by TimonReina et al. [10]. The graph database has a variety of uses, including network administration, social connectivity, biology, and the identification of fraudulent conduct. Compared to relational databases, it offers the developer more performance, flexibility, and agility [11]. Hybrid models and ensemble machine learning techniques also produce promising outcomes, with these techniques achieving results ranging from 87 to 97%. ROC, AUC, precision, recall, and F1 score are the performance measures used in the result analysis. This is mostly based on ROC curve analysis for each model [4]. Neural Network is another machine learning algorithm that when trained and tested for twenty epochs was able to give a mean squared error ( MSE) of 0.180 by Ghada and Abou-El-Enien. Metaheuristic Optimization algorithms, which help in increasing the prediction accuracy of machine learning algorithms [5]. The GTD codebook gives an overview of the data collection methodology for the global terrorism database. The code is maintained by the University of Maryland as well as it is copyright of the same. The database can be used on an individual basis for study purposes and is provided by the admin on a request basis [12].

3 Methodology 3.1 Graph Creation An application called Neo4j Sandbox was used to create the graphs. The global terrorism database was utilized as a source for building the graph and was imported into Neo4j by using the LOAD CSV cipher command. Using the CREATE command, several nodes were added to the graphs. Event Timing includes the date, time, and year of the occurrence; Event location includes the neighborhood, location, longitude, latitude, and Specificity, and attack types attacktype1, attacktype2, attack subtype 1, and attack subtype 2 were provided in the event info. Target type was composed of target types 1, 2, and target subtypes 1, and 2. Weapon type 1, weapon subtype 1, weapon type 2, weapon subtype 2, gun name, and gun type were all contained in the weapon type. Property damage, causality, and Ransome type were all the factors in causation. These nodes are all linked together through relationships. A cipher query language was used to generate and modify the graph. A subgraph of two nodes was constructed from the generated graph in order to compute graph embedding [7] (Fig. 2).

74

A. Raj et al.

Fig. 2 Sample graph created using Neo4j Sandbox

3.2 Graph Embedding Calculation Graph a higher dimension data is essentially represented in a lower dimension via embedding. They take the shape of a vector. In our example, embedding was calculated by using the node2vec technique. Node2vec algorithms operate based on random walks in the network. The graph is effectively represented in a lower dimension with a graph embedding by assuming a vector form [10]. The node2vec method was used to calculate embedding in our case. Node2vec methods use random network walks and are largely based on word2vec techniques. Fast random projection, node2vec, and Graphsage are the three techniques offered by Neo4j to compute node embeddings, node2vec is the approach used here. With the use of secondorder random walks, the node2vec method creates a list of node identities that, when put together, constitute a sentence. This corpus of sentences is then used to calculate embedding vectors, also known as node embeddings or graph embeddings. Based on random walks, the node2vec method alternates between depth-first search and breadth-first search [11]. Although up to 10 embedding dimensions have been generated in the research, the embeddings between two nodes in a network will be calculated based on the Neo4j platform up to n dimensions.

Terror Attack Classification with the Application of Orange Data …

75

4 Model Building The orange data mining tool was used to create many machine learning models. The global terrorism dataset and computed embedding are the first two input datasets that are entered into the orange tool by utilizing the CSV file import block. A Data table block may be used to visualize the CSV’s contents [2]. A Data table block may be used to visualize the CSV’s contents. The data table block’s output was provided as an input to the merge data block, which integrated the separate datasets into one dataset [3]. There are many null values in the merged dataset that can’t be directly given as input into the models and thus an imputation block was utilized to eliminate those null values. The select column block was incorporated into the models once null values were eliminated. The selected column block is used to choose and remove characteristics from the dataset as well as to choose and configure the model’s target variable. The purge domain block in the orange tool was used to delete and eliminate the redundant characteristics from the dataset since the data frame contains certain redundant attributes that were making the model’s prediction accuracy redundant. Figure 3 shows the employed model in the orange tool where X resembles the name of the ML models. Before feeding the dataset to the model, the dataset was scaled by using a data processor block. Overfitting is a severe problem that affects machine learning models most of the time. Here, principal component analysis (PCA) is used as a dimension reduction approach to solve this problem. The input data was then divided into

Fig. 3 Applied model in orange tool

76

A. Raj et al.

training and testing data by using an 8:2 sampling ratio. Eighty percent of the data were used to train the model, and twenty percent were used to test it, according to the sampling ratio of 8:2. The model block was the next to be added, and it received its input from the training database and its output from that block was provided as input to the test and score block. The test data table serves as the second input for the test and score block, which also assess how well the machine learning model performed. Multiple building elements, such as a bar plot, line plot, heat map, and others, can be used to visualize the results. In our instance, a confusion matrix was employed to visualize the outcome.

4.1 Counting Null Values The orange data mining tool’s impute block was used to count and eliminate null values from the dataset. There is a choice to use the average and most frequent imputation algorithm, random value imputation algorithm, model-based imputer method, or fixed value or numeric value imputation algorithm. In this study, the most frequent and average imputation procedure was used to remove null values from the dataset. The data imputation block for removing null values from the dataset is shown in Fig. 4.

Fig. 4 Data imputation block in orange

Terror Attack Classification with the Application of Orange Data …

77

4.2 Removal of Redundant Data The purge domain block in the orange data mining tool is used to eliminate or discard redundant characteristics from the dataset. Three alternatives are provided by the purge domain block to eliminate redundant characteristics from the dataset: features, classes, and meta attributes. The three major functions performed by purge domain block are sorting, reducing, and removing features. The purge domain block for removing redundant attributes from the dataset is shown in Fig. 5. Fig. 5 Purge domain block in orange tool

78

A. Raj et al.

Fig. 6 PCA in orange tool

4.3 Dimension Reduction Overfitting, which happens as a result of the dataset’s many characteristics, is one of the main issues that machine learning models encounter. There are several methods for reducing the number of dimensions, including PCA (Principal Component Analysis) and LDA (Linear Discriminant Analysis). PCA (Principal Component Analysis) is used in this study to reduce the dimensions to two components. With a variance of 99%, the PCA block in the orange data mining tool is utilized for performing dimension reduction. The PCA block for dimension reduction from the dataset is shown in Fig. 6.

4.4 Data Scaling The process of bringing the data into a certain range so the model can quickly learn and categorize is known as data scaling. The orange tool has a number of scaling techniques, including conventional scaling and center scaling. For data scaling, utilize the orange data processor block. Data preparation options available in the data preprocessor block include discretization, continuization, imputation, normalizing, randomization, and principal component analysis. The data processor block for data scaling from the dataset is shown in Fig. 7.

Terror Attack Classification with the Application of Orange Data …

79

Fig. 7 Preprocess block in orange tool

5 Results and Analysis The obtained research findings indicate that Random Forest (RF) with accuracy scores of 0.938, F1 scores of 0.920, precision scores of 0.936, and recall scores of 0.932 is the model that performs the best. SVM is the model that performs the poorest, with accuracy scores of 0.554, F1 scores of 0.830, precision scores of 0.850, and recall scores of 0.880. AUC, precision, recall, and F1 Score are the study’s performance indicators. Table 1 shows the results obtained from the proposed research. Table 1 Results obtained from the proposed research Model

AUC

F1 score

Precision

Recall

RF

0.938

0.920

0.936

0.932

GB

0.931

0.955

0.954

0.955

KNN

0.743

0.850

0.842

0.870

Tree

0.572

0.903

0.902

0.911

SVM

0.554

0.830

0.850

0.880

NB

0.833

0.710

0.869

0.651

Adaboost

0.862

0.927

0.931

0.925

80

A. Raj et al.

To determine accuracy, apply the formula below:  Accuracy = (T P + T N ) (T P + T N + F P + F N ) To determine the precision, apply the formula below:  Pr ecision = T P (T P + F P) To determine the recall, apply the formula below:  Recall = T P (T P + F N ) To determine the F1 Score, apply the formula below:  F1 Scor e = 2 ∗ (Pr ecision ∗ Recall) Pr ecision + Recall where TP is the true positive classified sample by the model, TN is the true negative classified sample by the model, and FN is the false positive classified sample by the model, FN is the false negative classified sample by the model. Figure 8 shows the confusion matrix for the random forest model. Figure 8 shows the confusion matrix for the random forest model. Figure 9 shows the confusion matrix for the gradient boosting model. Figure 10 shows the confusion matrix for the KNN model. Figure 11 shows the confusion matrix for the decision tree model. Figure 12 shows the tree diagram for the classification model. Figure 13 shows the confusion matrix for the SVM model. Figure 14 shows the confusion matrix for the naive bayes model. Figure 15 shows the confusion matrix for the Adaboost model. Fig. 8 Confusion matrix for random forest

Terror Attack Classification with the Application of Orange Data …

81

Fig. 9 Confusion matrix for gradient boosting

Fig. 10 Confusion matrix for KNN

Fig. 11 Confusion matrix for decision tree

6 Conclusion Tools like Orange and WEKA can be very useful in the absence of knowledge on python programming language. Graph features such as graph embedding can act as useful features in the classification process. Random Forest (RF) and Gradient

82

Fig. 12 Tree diagram for decision tree Fig. 13 Confusion matrix for SVM

Fig. 14 Confusion matrix for naive bayes

A. Raj et al.

Terror Attack Classification with the Application of Orange Data …

83

Fig. 15 Confusion matrix for Adaboost

Boosting (GB) techniques are the most promising techniques used for the classification of terror attacks by using the orange data mining tool. According to our findings, Random Forest (RF) can provide an accuracy score of 0.938, F1 score of 0.920, precision of 0.936, and recall of 0.932 on a training and testing ratio of 8:2. The worst performing model was SVM, which gave an accuracy score of 0.554, F1 score of 0.830, precision of 0.850, and recall of 0.880. Although orange is a very diverse tool with its own limitation including less flexibility, a predefined and limited number of algorithms, and less customization of blocks. Graph embedding has compensated the inconsistency in the dataset and improves the prediction accuracy of the model.

References 1. Lopez FMS, De La Cruz EGS (2015) Literature review about Neo4j graph database as a feasible alternative for replacing RDBMS. Revista de la Facultad de Ingenieria Industrial 1560–9146 2. Peker M, Özkaraca O, Sa¸ ¸ sar A (2018) Use of orange data mining toolbox for data analysis in clinical decision making: the diagnosis of diabetes disease. In: Expert system techniques in biomedical science practice 3. Ranjan R, Agarwal S, Venkatesan S (2017) Detailed analysis of data mining tools. Int J Eng Res Technol (IJERT) 2278–0181 4. Python A, Bender A, Nandi AK, Hancock PA, Arambepola R, Brandsch J, Lucas TCD (2021) Predicting non-state terrorism worldwide. Sci Adv 5. Soliman GMA, Abou-El-Enien THM (2019) Terrorism prediction using artificial neural network. Revue d’Intelligence Artificielle 6. Macák M, Stovcik M, Buhnova B (2020) The suitability of graph databases for big data analysis: a benchmark. IoTBDS 7. Tolan GM, Soliman OS (2015) An experimental study of classification algorithms for terrorism prediction. Int J Knowl Eng 8. Huamaní EL, Alicia AM, Roman-Gonzalez A (2020) Machine learning techniques to visualize and predict terrorist attacks worldwide using the global terrorism database. Int J Adv Comput Sci Appl (IJACSA)

84

A. Raj et al.

9. Pokorný J (2015) Graph databases: their power and limitations. In: Computer ınformation systems and ındustrial management 10. Timon-Reina´ S, Rincon M, Martínez-Tomas R (2021) An overview of graph databases and their applications in the biomedical domain. Database J Biol Database Curation 11. ShefaliPatil G, Bhatia A (2014) Graph databases—an overview. Int J Comput Sci Inf Technol 12. Global terroism database. https://www.start.umd.edu/gtd/downloads/Codebook.pdf

Multipurpose IoT Based Camera Using Deep Learning Urvashi Dube, Sudhish Subramaniam, and G. Sumathi

Abstract The COVID 19 pandemic has given rise to a new normal. This includes wearing masks and maintaining social distance. Nowadays sudents don’t focus in offline classes. Also, students with masks in offline proctored exams find ways to roll their eyes at others’ work for malpractice. The systems designed to date are not accurate to detect facial features with mask. These problems have motivated us to develop a reliable, robust model to detect mask, eye location, eyeball location, eye status, and head pose of people wearing and not wearing a mask, all at once. We have used 3800 masked, unmasked images to train our model using MobileNetV2, a convolutional neural network, with 99% accuracy. The output of this model is processed using image processing, facial landmark analysis, EAR, and deep learning to detect the facial landmarks accurately. Ultimately, a unique method is used to detect head pose of person. Keywords Image processing · Deep learning · Face mask · Eye location · Eyeball location · Head pose estimation

1 Introduction Covid 19 pandemic has changed the style of living in the world. Although the virus is exiting, it has an impact on people. Schools and colleges have opened but people still wear masks to all places. In this masked environment, face image processing has faced a lot of challenges. To overcome all these problems, we have developed a model to detect the face mask accurately and to detect the eye, eye status, eye location, eyeball location and the head pose of the person in front of the camera. Our model is also designed to detect the attentiveness of the person in a classroom. This is one of the major problems faced in the school environment, students are not paying attention in the classroom under the masks, and this decreases the productivity of education in the schools, colleges and classes. Our model can also be used in the U. Dube · S. Subramaniam · G. Sumathi (B) School of Electronics, VIT, Vellore, Tamil Nadu, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_7

85

86

U. Dube et al.

offline proctored exams where students are not allowed to roll their eyes elsewhere during the duration of the exam. To avoid malpractice, our model can be used in runtime to capture images and process the eye landmarks and head pose of the candidate. The model can be further progressed to end the test of the user if the person has attempted malpractice. We have used mobilenet_v2 to classify the 3800 images of masked and unmasked people. On getting a very high accuracy of 99%, we proceeded to filter out the eyes and other facial features. We ultimately determine mask, eye status, eye location, eyeball location, and head pose irrespective of the person wearing or not wearing a mask.

2 Related Work The masking of a face can be identified by an edge computing-based deep learning algorithm [1], this method is specifically implemented in busses to identify people wearing masks or not. A dataset of masked faces (MAFA) and two CNN algorithms (LLE-CNN algorithm) are applied in [2]. A model comprising of PIR Sensor, microcontroller and smartphone system is proposed in [3] to detect the motion and store the corresponding output video in the cloud. In [4], a generative adversarial network for masked object detection and image completion of the removed masked region is proposed and implemented. Detection of wearing state of face mask by training a custom dataset: a face without a mask, face with the wrong mask, face with the correct mask is done using Context-Attention R-CNN method in [5]. Head Pose estimation can also be done using the HGL method which is a combination of the H-channel of the HSV colour space with the face portrait and grayscale image, this method is proposed in [6] and achieves an accuracy of 87.17%. The paper [7] includes two novel ideas, a residual context attention module for crucial face mask related regions and an auxiliary task using a synthesized gaussian heat map regression method to discriminate features of the face. The authors of [8] have proposed a model to detect candidates wearing mask regions using the transfer model of Faster_RCNN and InceptionV2 structure, in the second stage real facial masks are verified using a broad learning system. A simple and effective facial landmark detection method comprising of a lightweight U-Net model and a dynamic optical flow is proposed in [9] which exhibits better performance than others without requiring heavy computational loadings. The authors of [10] have developed a model using driving environment datasets and eye aspect ratio (EAR), to detect facial landmarks, eye location and state evaluation, they achieved an average accuracy of 93.9%. The authors of [11] used the method of segmentation of pupil and iris images by pixel to determine the eye status of the driver and his fatigue, they achieved an accuracy of 96.72%. In [12] fatigue detection convolutional network (FDCN) based CNN network was built which has a 1.0% accuracy improvement on the ZJU database on fatigue detection. DriCare, a new face-tracking algorithm to improve the tracking accuracy is developed in [13], and it achieved 92% accuracy. Eye status, PERCLOS of both coloured and infrared, fatigue is detected around the clock in [14]. Using logistic regression,

Multipurpose IoT Based Camera Using Deep Learning

87

EAR and analyzing facial landmarks percent eye-closure over a fixed time window (PERCLOS), blink rate, statistics of blink duration, closing speed, reopening speed and number of yawns are extracted in [15]. In [16], eyes for a frontal face are extracted precisely. Histograms of Oriented Gradient (HOG) descriptors are proved to be better than existing models in the case of human detection in [17]. You Only Look Once (YOLO) method for object detection uses regression models instead of repurposing classifiers is developed in [18], this method outperforms DPM, R-CNN methods. ˙In [19] overall face detection, facial feature localization, and face comparison is carried out all at once. The authors of [20] have built a model to gather environmental parameters to build a smart campus environment. The parameters include air temperature, light intensity, and humidity. The paper also carries out an in-depth study on how to store real-time data in a standard and organized manner. The work in [21] is about principal component analysis (PCA) used for feature extraction that helps in achieving superior performance. The authors of the paper have worked to achieve a high recognition rate for IoT based image recognition. On analyzing all the pre-existing researches, it is evident that a robust model to detect all the facial features and head pose of a person with a mask all at once, does not exist. This has motivated us to develop a plentiful model which extracts all the features mentioned, from an image.

3 Proposed Work 3.1 Hardware Integration To test the model’s performance with different cameras and different lighting, we have used a raspberry pi camera, webcam, CCTV security camera and laptop webcam. The raspberry pi board can be integrated easily with Rpi camera and can also be used to transfer images from one computer to the other, in the same network using file transfer protocol (FTP). The raspberry pi has a very fast processing speed and capability to run long codes as compared to other boards. Using Rpi OS, the Raspberry pi provides an interface to work on and can be programmed using python, C, etc. The raspberry pi camera or a portable webcam, is interfaced with the raspberry pi. Using a python code, the images are sent periodically through file transfer protocol to the Jupyter notebook for image processing. In the case of the raspberry pi camera, picamera module is used to interface the camera with the raspberry pi. Imwrite function of OpenCV module is used to store the images on the RPI OS. File Transfer Protocol is a set of guidelines that controls how computers transfer data across the internet from one system to another. An FTP server is first set up, by connecting the raspberry pi and laptop to the same Wi-Fi network as shown in Fig. 1. ftplib library of python is used to send images through FTP. In the case of CCTV cameras, the TAPO camera serves Real-Time Streaming Protocol (RTSP). The protocol combines intricate programming, transcoding and client server method to send video through a

88

U. Dube et al.

Fig. 1 Using FTP folder with webcam and RPI

network or to the internet using a link. Using this protocol, we have written a python code to display the stream on the Jupyter notebook. By analyzing the frames per second (FPS) we extract the frames periodically and save them in the OS using the OpenCV library. These images are then used for image processing and the results are obtained. In the case of a laptop webcam, images are captured and directly sent to the respective folder where images are extracted one at a time for image processing. The flowchart of working of hardware components of the model is shown in Fig. 2a–c.

3.2 Training the Model Initially we import ImageDataGenerator, MobileNeyV2, AveragePooling2D, Dropout, Flatten, Dense, Input, Model, Adam, preprocess_input, img_to_array, load_img, to_categorical libraries from their respective Tensorflow.keras libraries. For preprocessing the dataset, we imported libraries from Sklearn. Other imported libraries included utils, matplotlib, NumPy, argparse and OS. Then we initialize

Multipurpose IoT Based Camera Using Deep Learning

89

Fig. 2 a Flowchart of working of model with CCTV camera and Jupyter notebook, b flowchart of working of model with laptop webcam and Jupyter notebook, c flowchart of working of model with Rpi and Rpi camera

the initial learning rate, number of epochs to train for, and batch size which are determined through the hit and trial method. Once the dataset (with 3800 images) is imported, the model loops over all the images in the folder by simultaneously labelling them. Data and labels are converted into NumPy arrays and passed to a LabelBinarizer to fit the model. The dataset is split into train and test with 80% for training and 20% for testing. Data augmentation is also applied to increase the accuracy. After creating the base model with the help of MobileNetV2, head Model is made using the layers AveragePooling2D, Flatten, Dense, Dropout and again Dense layer. The head model is then placed on top of the base model. Finally, the model is fit with 20 epochs. The model is tested with the testing dataset and we get the classification report. Using this method, we have achieved an accuracy of 99% for our model.

3.3 Detect and Predict Mask The extracted frames/images are sent as arguments to this function. FaceNet and MaskNet have defined weights for the convolutional neural network applied. Once the frame/image is obtained, first the height and weight are extracted and saved in shape. A blob image is made using the weights from faceNet. FaceNet contains a module, forward, that detects frontal faces, this is applied to the captured image. If

90

U. Dube et al.

Fig. 3 Input image (left), output image (right), mask applied on the input image to detect eyes (white portion detecting eyes). Source Adapted from [21]

the confidence of the detected face is greater than 50%, the coordinates of the face are extracted and saved in locs. If a face has been detected, maskNet is used to detect the face with a mask on it. Once this is predicted, the coordinates are saved in preds. Ultimately a tuple of locs and preds is returned.

3.4 Eye on Mask To detect the eyes on a face, we have applied a mask on the whole face except on the eyes. This helps to detect the eye location accurately. For this function, the predefined mask and the side of the eyes are sent as arguments. The location of the predefined eyes is saved in points, it is then converted into a NumPy array for processing. Once all the preprocessing is done, the fillConvexPoly () function is used to fill the face with a mask except for the eyes as shown in Fig. 3.

3.5 Eye Open or Closed We have used the Eye Aspect Ratio (EAR) method to determine the status of the eye. In this method, once the eye location is obtained, the distance between the upper and lower eyelid is calculated. The points used to calculate EAR are shown in Fig. 4 (red points). This is compared with the standard value of eyes open and closed. It helps

Multipurpose IoT Based Camera Using Deep Learning

91

Fig. 4 Pictorial image of an eye marked red points used to calculate EAR

in determining if the eye is closed or opened. The standard values for a masked and unmasked face are different, the functions are defined respectively.

3.6 Eye and Eye Ball Contours In the case of eye contours, the shape_68.dat is used to extract the coordinates of the eye. Once the coordinates are obtained it is sent as arguments to the eye contour function. The circle() function of OpenCV module takes five arguments, including the source picture, the (x, y) coordinates, the radius, the colour, and the thickness. The circle function draws a circle on the coordinates of the image with the mentioned radius and thickness. Obtained coordinates, radius = 2, respective color and thickness = 2 are passed to the function to draw the contours. In the case of the eyeball contours, a threshold value is set (different for masked and unmasked faces), the midpoint of the eyeball is calculated and the threshold, midpoint and image are sent with right = False. Now, each eye is taken at a time and a mask is applied to them. Once the mask is applied, contours are detected. These contours are found by adjusting the threshold values sent. The contours detected are eyeballs. Obtained coordinates are marked using the circle function of OpenCV.

3.7 Head Pose of Person The image obtained from hardware is sent as an argument to the function. The image is first resized to 1000 × 600 pixels and then flipped for processing. Then the image is converted to grayscale, a face mesh is created on the detected image. Using the defined weights, the landmarks of the face are detected. Once detected a loop is iterated on the coordinates. Using the nose coordinates and the other coordinates of the face, rotational and translational vectors are created using the function solvePnP of OpenCV. The rotational vector is then sent to rodrigues function of OpenCV to calculate rmat and jac. Rmat is then sent to RQDecomp 3 × 3 to get the angles (x, y, z) that are roll, pitch and yaw of the face. While testing the algorithm for different images, we found that the images form a pattern, when (x, y, z) is summed and compared. For different images, the summation of roll, pitch and yaw values had

92

U. Dube et al.

different outputs. Hence the method is generalized and it became a distinguishing factor in the head pose estimation. Finally, the summation of the angles is returned.

3.8 Function Calling and Display of Output On reading the image, it is converted into grayscale and locs and preds are found. We predict if the person has worn a mask or not. On detecting a mask, we find the locs and preds to create a rectangle in which we detect the eye contours and eyeball contours, eye status is found using Eye Aspect Ratio. Ultimately the head pose of the person is found. The output is displayed on the image. Then we display the final output image with all the labels. If the face is unmasked, we use the detector to find the rectangle surrounding the face. Then we find the eye status, eye contours, eyeball contours and head pose. The final output image is displayed with all the labels. The flowchart in Fig. 5 illustrates the process.

Fig. 5 Flowchart of application of the developed algorithm

Multipurpose IoT Based Camera Using Deep Learning

93

4 Result Analysis We used a laptop webcam for the input image. The input and output images are shown in Fig. 6. Part 2: Interfacing raspberry pi camera and laptop using File Transfer Protocol (FTP). The results of image processing are shown in Fig. 7. Part 3: Interfacing webcam, raspberry pi and laptop for the input image we received accurate results. The processed images are shown in Fig. 8. Part 4: Interfacing Tapo camera directly with Jupyter notebook by extracting the stream using Real-Time Streaming Protocol (RTSP). We saved the frames of the stream at regular intervals. The images and their respective processed output images are shown in Fig. 9. Part 5: Combining all parts, the head pose of the person is also displayed on the image as shown in Fig. 10. The training loss and accuracy versus epoch plot are shown in Fig. 13. Loss and accuracy functions from the deep Learning model, training history, in keras is used to find the training loss and accuracy of the model. It is clearly evident that the accuracy of the model is 99% and the training loss of the model is 1%, when the number of epochs reaches 20. The epoch results with 20 epochs are shown in Fig. 14. The classification report with precision, recall, f1-score and support is shown in Fig. 15. The accuracy of the model increases and validation loss of the model decreases with

Fig. 6 Input image from a laptop webcam

94

U. Dube et al.

Fig. 7 Using a raspberry pi camera to get input image for masked and unmasked image

Fig. 8 Using webcam and Rpi to get input image for masked image

Multipurpose IoT Based Camera Using Deep Learning

95

Fig. 9 Using Tapo camera to get input image

Fig. 10 Final output images with all the features and head pose

increasing epochs. Once the model is trained, it is fit into the input images to get the output. The model achieves 99% accuracy with both mask images and unmask images. The obtained accuracy is better than other existing models. The accuracy is calculated using the below formula: Accuracy =

T r ue categor y1 + T r ue categor y2 T r ue categor y1 + T r ue categor y2 + False categor y1 + False categor y2

96

U. Dube et al.

Fig. 11 Final output images with all the features and head pose (left)

Computationally, the training loss is calculated by taking the sum of errors for each example in the training set. The training loss is measured after each batch. The results obtained are accurate and the model is working perfectly. It can be described as a robust model. The method of detecting each landmark one after the other has increased the performance of the model. The results from the laptop webcam are shown in Figs. 11 and 12. Here the image processing is done with a grayscale image, this increases the performance of the model too. The rotational vector calculated during the head pose estimation differs from one camera to another as the focal length of each camera is different. The model works best when the image processing is done at regular intervals, this reduces the processor requirements. Images are sent at regular intervals to the Jupyter notebook for image processing. Once the image processing is completed, the output image with labels is displayed. The model can be used to alert the alarm system if a person, wearing or not wearing a mask, is not attentive in a class or in a driving system. In a classroom, once the direction of the teacher or the blackboard is set, the model can be modified to detect if the pupil is attentive in the class or not. The major limitation of the model is that it cannot be used in very dim light conditions or at night, where there is minimum or no light. To overcome this problem, night vision cameras can be incorporated, to get the input stream, which can be processed by the model to get the desired results. Another limitation is the hardware interconnections between the camera and the raspberry are prone to wear and tear, and have to be handled with care for accurate results and smooth processing.

5 Conclusion We have achieved a maximum accuracy of 99% and our model is working better than the pre-existing models. We have also worked on head pose estimation and we achieved better accuracy and precision than the existing solutions. Our model is best

Multipurpose IoT Based Camera Using Deep Learning

Fig. 12 Final output images with all the features and head pose (right)

Fig. 13 Training loss and accuracy versus epoch plot

Fig. 14 Epoch results of the trained model

97

98

U. Dube et al.

Fig. 15 Classification report of fit model

fit in real-time applications and can be used in different scenarios in driving environments, schools, colleges, and offline proctored exams to examine the attentiveness of a person accurately. The model can be trained and modified to get appropriate results in the night too. Night vision cameras can be incorporated in the model to get images in the night. The night vision cameras could be fit inside a car, where the lighting conditions are very less or negligible, to check the status, eye status and head pose, of a masked driver. In case of any abnormality in the driver status or negligence in driving an alarm could be triggered to aware about a possible accident. Our developed model outperforms the pre-existing models, faster_RCNN, inceptionV2 structure, fatigue detection convolutional network (FDCN), Histograms of Oriented Gradient (HOG), of image processing to detect if a person has worn a mask or not or to detect facial features without a mask. The achieved accuracy of the model makes it robust and fit for all lighting conditions and angles.

References 1. Kong X et al (2021) Real-time mask identification for COVID-19: an edge-computing-based deep learning framework. IEEE Internet of Things J 8(21):15929–15938. https://doi.org/10. 1109/JIOT.2021.3051844 2. Ge S, Li J, Ye Q, Luo Z (2017) Detecting masked faces in the wild with LLE-CNNs. In: 2017 IEEE conference on computer vision and pattern recognition (CVPR), pp 426–434. https://doi. org/10.1109/CVPR.2017.53 3. Sanoob AH, Roselin J, Latha P (2016) Smartphone enabled intelligent surveillance system. IEEE Sens J 16(5):1361–1367. https://doi.org/10.1109/JSEN.2015.2501407 4. Din NU, Javed K, Bae S, Yi J (2020) A novel GAN-based network for unmasking of masked face. IEEE Access 8:44276–44287. https://doi.org/10.1109/ACCESS.2020.2977386 5. Zhang J, Han F, Chun Y, Chen W (2021) A novel detection framework about conditions of wearing face mask for helping control the spread of COVID-19. IEEE Access 9:42975–42984. https://doi.org/10.1109/ACCESS.2021.3066538 6. Li S et al. (2020) Multi-angle head pose classification when wearing the mask for face recognition under the COVID-19 coronavirus epidemic. In: 2020 International conference on high performance big data and intelligent systems (HPBD&IS), pp 1–5. https://doi.org/10.1109/ HPBDIS49115.2020.9130585 7. Fan X, Jiang M, Yan H (2021) A deep learning based light-weight face mask detector with residual context attention and Gaussian heatmap to fight against COVID-19. IEEE Access 9:96964–96974. https://doi.org/10.1109/ACCESS.2021.3095191

Multipurpose IoT Based Camera Using Deep Learning

99

8. Wang B, Zhao Y, Chen CLP (2021) Hybrid transfer learning and broad learning system for wearing mask detection in the COVID-19 Era. IEEE Trans Instrum Meas 70:1–12, Art no. 5009612. https://doi.org/10.1109/TIM.2021.3069844 9. Wu B-F, Chen B-R, Hsu C-F (2021) Design of a facial landmark detection system using a dynamic optical flow approach. IEEE Access 9:68737–68745. https://doi.org/10.1109/ACC ESS.2021.3077479 10. Ling Y, Luo R, Dong X, Weng X (2021) Driver Eye location and state estimation based on a robust model and data augmentation. IEEE Access 9:67219–67231. https://doi.org/10.1109/ ACCESS.2021.3076365 11. Zhang F, Su J, Geng L, Xiao Z (2017) Driver fatigue detection based on eye state recognition. In: Proceedings of the international conference on machine vision and information technology (CMVIT), pp 105–110 12. Huang R, Wang Y, Guo L (2018) P-FDCN based eye state analysis for fatigue detection. In: Proceedings of the IEEE 18th international conference on communication technology (ICCT), pp 1174–1178 13. Deng W, Wu R (2019) Real-time driver-drowsiness detection system using facial features. IEEE Access 7:118727–118738 14. Sun Y, Yan P, Li Z, Zou J, Hong D (2020) Driver fatigue detection system based on colored and infrared eye features fusion. Comput Mater Continua 63(3):1563–1574 15. Cheng Q, Wang W, Jiang X, Hou S, Qin Y (2019) Assessment of driver mental fatigue using facial landmarks. IEEE Access 7:150423–150434 16. Colak ME, Varol A (2020) Easymatch-an eye localization method for frontal face images using facial landmarks. Tehniˇcki Vjesnik 27(1):205–212 17. Dalal N, Triggs B (2005) Histograms of oriented gradients for human detection. In: Proceedings of the IEEE computer society conference on computer vision and pattern recognition (CVPR), pp 886–893 18. Redmon J, Divvala S, Girshick R, Farhadi A (2016) You only look once: unified, real-time object detection. In: Proceedings of the IEEE computer society conference on computer vision and pattern recognition (CVPR), pp 779–788 19. Colmenarez AJ, Huang TS (1998) Face detection and recognition. In: Face recognition. Springer, Berlin, Heidelberg, pp 174–185 20. Sungheetha A (2021) Assimilation of IoT sensors for data visualization in a smart campus environment. J Ubiquit Comput Commun Technol 3(4):241 21. Jeena JI, Darney PE (2021) Design of deep learning algorithm for IoT application by image based recognition. J ISMAC 3(03):276–290

Dr. Watson AI Based Healthcare Technology Project N. Suresh Kumar, S. Ganesh Karthick, K. P. Aswin Kumar, S. Balaji, and T. Nandha Sastha

Abstract The aim of the design is to implement Artificial Intelligence in the Healthcare domain and find suitable results. This work is an AI based web application which enables four features 1. Medicine Prescriber 2. Diabetic Analyzer 3. Covid19 Predictor and 4. AI based chatbot. Provide online solution for the patients like prescribing medicine based on the symptoms, Analyzing the blood sugar (Mg/Dl) and suggesting food diet based on the age and sugar level, Predicting Covid-19 Positive or negative according to the X-ray of the chest and AI based Chatbot which acts as a support agent guides the user in using this application and tells interesting facts on Covid-19. The UI design of the web application is crafted using Adobe XD. Machine learning and Deep learning techniques are used to predict results for these features, these Machine learning and Deep Learning models are deployed as Web application using a framework called Flask. IBM Watson Assistant which is used to create the chatbot, allows you to integrate conversational interfaces into any app, device, or medium, as well as add a natural language interface to the app to automating conversations with your customers. Keywords Dr. Watson AI based healthcare technology · Medicine prescriber · Covid predictor · Diabetic analyser · Chatbot

N. Suresh Kumar Department of IT, Sri Ramakrishna Engineering College, Coimbatore, Tamil Nadu, India N. Suresh Kumar · S. Ganesh Karthick (B) · K. P. Aswin Kumar · S. Balaji · T. Nandha Sastha Sri Ramakrishna Engineering College, Coimbatore, Tamil Nadu, India e-mail: [email protected] K. P. Aswin Kumar e-mail: [email protected] S. Balaji e-mail: [email protected] T. Nandha Sastha e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_8

101

102

N. Suresh Kumar et al.

1 Introduction Patients, doctors, and hospital managers’ life are made easier by artificial intelligence, which performs activities normally performed by people in a relatively short time and at a fraction of the investment [1]. The AI sector, which was estimated at around $600 million in 2014 and is expected to hit $150 billion by 2026, is one of the world’s fastest-growing sectors [2]. In the simplest sense, AI is something that whenever computers as well as any other machines can study, analyze, and decide things or take actions in the same way that humans can [3]. The use of machines to evaluate and act on medical information, usually with the purpose of predicting a specific outcome, is referred to as AI in healthcare [4]. AI in healthcare can improve patient outcomes through improving preventative care and livability, as well as producing more accurate diagnosis and treatment regimens. By data analysis from the government, hospitals, as well as other resources, AI can also anticipate and track the spread of contagious diseases [5]. As a result, AI has the potential to be a critical instrument in the fight against diseases and pandemics in global public health. Artificial intelligence in medicine uses Machine learning algorithms to search medical data for ideas that can assist increase health and treatment outcomes [6]. Clinical decision assistance and image analysis are now the most prominent uses of AI in medical contexts [7]. Clinical decision support tools assist practitioners in making therapy, medicine, mental health, and other patient-related decisions by providing quick access to relevant information or research. AI, unlike humans, does not require sleep [8]. Machine learning algorithms could be used to monitor the vital signs of critically ill patients and notify clinicians if specific risk indicators rise. Artificial Intelligence is used to spot patterns in behaviour that lead to either high or low blood sugar levels in diabetes patients [9]. Continuous glucose monitors used by those with diabetes collect a huge amount of data that has previously not been used efficiently. Type 2 diabetes is observed to be more frequent than type 1. Type 1 diabetes affects more than one out of every ten persons, whereas type 2 affects the rest. With age, the percentage of diabetics rises. Diabetes affects approximately 10.5% of the general population. In the 65 and older age categories, the rate is as high as 27%. Here the AI analyse the blood sugar value and personalized healthcare recommendations like food diet. This project also focusses on predicting Covid-19 positive or negative according to the X-ray soft copy of the chest, this prediction is made based on the Deep learning technique where large amount of chest X-ray images dataset of soft copy of covid positive and negative images are trained [10]. An AI based virtual chatbot is developed using IBM Watson Assistant is a white label cloud service that enables business software developers to integrate an artificial intelligence virtual assistant into their product and brand it as their own. The IBM Cloud is used to deliver the service, which allows customers access to Watson AI. Here this chatbot is developed to tell facts on Covid-19 and guides the user in using this Web application.

Dr. Watson AI Based Healthcare Technology Project

103

2 Methodology The entire Web application runs on a Framework called Flask, where this application contains four features. 1. Medicine Prescriber 2. Diabetic Analyzer They have a unique dataset for each feature and Machine Learning Models are created for each feature with their respective dataset. Here the libraries used to train the machine learning model are Pandas, NumPy, SciKit, Pickle. This Machine Learning model is created using Decision Tree classifier which by constructing a decision tree, the categorization model is created. Every node in the tree represents a test on a variable, and each branch falling from that node represents one of the property’s possible values. Each branch represents one of the instance’s class labels. The training set’s instances are identified by routing them from the base of the tree to a leaf, based on the results of the tests all along way. Each node in the tree divides the instance space into consists of two subs based on an attribute test condition, starting with the root node. After that, a new node is produced by moving down the tree branch matching to the attribute’s value. This cycle continues for the subtree rooted just at new node, until the training set’s records have all been classified. The decision tree is normally built from the top down, with each step selecting the optimal attribute test condition for splitting the data. There are a variety of methods for determining the optimal reason to divide the records. 3. Covid Predictor, a large amount of chest X ray is used as dataset and trained using Convolutional Neural Networks (CNN) are plug neural networks that are used to assess visual images by data processing in a grid-like layout. CNN is selected to train a deep learning model since it trains the input images into many hidden layers and brings out the trained classified output. A ConvNet is another name for it. To discover and classify items in an image, a convolutional neural network is employed. An accuracy of 88.7% is attained in this deep learning model. This Deep Learning model is created using TensorFlow, Keras and OpenCV libraries modules. Pickle is a library which is used to save the machine learning and deep learning model and deploy those saved model as a web application using the framework called Flask. Flask is a python scripted language microframework for creating small websites, and creating Restful APIs in Python is quite simple. The UI of the web application is crafted using Adobe XD and built as a web application using front end frameworks like HTML, CSS, JavaScript, Bootstrap.

3 Implementation The home page which contains the Logo and header menus, and a chatbot below. To view the features user, need to click Meet Dr. Watson! Button. The chatbot which helps the user to guide how to use these features and get interesting facts on covid-19. Dashboard screen is a graphic representation of all of the characteristics. Although it can be utilized in a variety of ways, its primary goal is to enable quick access

104

N. Suresh Kumar et al.

Fig. 1 Home page

to information. This enables user to view three features 1. Medicine prescriber 2. Diabetic Analyzer 3. Covid Predictor. Medicine prescriber feature allow user to enter their symptom and get medicine prescription. Diabetic analyser page which helps user in analysing their diabetic and get proper food diet based on their age and sugar level (mg/dl). The user can also download the result for future purpose. Covid predictor feature predicts covid positive or negative based on the X-ray soft copy of chest. User is required to upload the soft copy of the chest X-ray and click submit to view the result. Team is a page where the team members names are mentioned in. to navigate to team simply click team button in header. Technology is a page where the technology used in this project are mentioned in a descriptive manner. To navigate to this page user simply need to click the Technology button in the header.

4 Results 4.1 Home Page Select “Meet Dr. Watson” button and enter the Dashboard page (Fig. 1).

4.2 Dashboard Page Dashboard screen will help the user to get directed to the features (Fig. 2).

Dr. Watson AI Based Healthcare Technology Project

105

Fig. 2 Dashboard page

4.3 Medicine Prescriber This feature prescribes medicine based on the user symptoms. The decision tree classifier algorithm predicts the best suitable medicine for the symptoms and allow the user to download and print the medicine as a medical prescription (Fig. 3).

Fig. 3 Medicine prescriber

106

N. Suresh Kumar et al.

4.4 Diabetic Analyzer This feature analyses the user input which is their age and blood sugar level (mg/dl) of before food and after food and make predictions using the machine learning algorithm and provide the diabetic condition status and suggest proper food diet based on the user input. The user can download and print that food diet to kept a note of it (Fig. 4).

4.5 Covid Predictor This feature allows user to upload the chest X-ray softcopy and check whether Covid positive or negative, when the softcopy of the chest X-ray scan which is provided as a user input the Convolutional Neural Network model predicts best suitable results. An accuracy of 88.7% is attained in this work, enables user to experience a fast response (Fig. 5).

4.6 Watson Assistant Chatbot This feature guides the user to use this web application and acts as a support agent for this application and also capable of answering questions related to Covid-19. These Covid facts are trained based on the World Health Organization (WHO) data using Natural Language Processing (NLP) (Fig. 6).

5 Conclusion Health care is progressively expanding into the house, involving a mix of individuals, a variety of jobs, and a wide range of instruments and technologies; it also occurs in a variety of residential settings. Rising health-care costs, rising numbers of seniors, rising prevalence of chronic disease, improved sustenance rates of various illnesses, injuries, as well as other conditions, massive groups of returning veterans from war with serious injuries, and a broad range of new technologies are all driving this migration. The quality and cost of the health care provided as a result varies greatly in terms of safety, efficiency, and efficiency as well as performance and price.

Dr. Watson AI Based Healthcare Technology Project

Fig. 4 Diabetic analyzer

107

108

Fig. 5 Covid predictor

Fig. 6 Watson assistant chatbot

N. Suresh Kumar et al.

Dr. Watson AI Based Healthcare Technology Project

109

References 1. Chen X, Xie H, Zou D, Hwang GJ (2020) Application and theory gaps during the rise of artificial intelligence in education. Comput Educ Artif Intell 1:100002 2. Kandlhofer M, Steinbauer G, Hirschmugl-Gaisch S, Huber P (2016) Artificial intelligence and computer science in education: from kindergarten to university 3. Tobore I, Li J, Yuhang L et al. (2019) Deep learning intervention for health care challenges: some biomedical domain considerations 4. Rathore PS, Sharma BK (2022) Improving healthcare delivery system using business intelligence. J IoT Soc Mob Anal Cloud 4(1):11–23 5. Mason A, Morrison A, Visintini S (2018) An overview of clinical applications of artificial intelligence 6. Rajkomar A, Dean J, Kohane I (2019) Machine learning in medicine. N Engl J Med 7. Gajane P, Pechenizkiy M (2017) On formalizing fairness in prediction with machine learning 8. Adamson AS, Smith A (2018) Machine learning and health care disparities in dermatology 9. Collins GS, Moons KGM (2019) Reporting of artificial intelligence prediction models 10. Smys S (2019) Survey on accuracy of predictive big data analytics in healthcare. J Inf Technol 1(02):77–86

Empirical and Statistical Comparison of RSA and El-Gamal in Terms of Time Complexity Ankita Kumari, Prashant Pranav, Sandip Dutta, and Soubhik Chakraborty

Abstract In this paper, two algorithms are compared based on their time complexity. The time complexity is defined by encryption and decryption of different message lengths. Time varies for different lengths of messages. We statistically analyzed the time complexity of the algorithm and compared their results. Keywords RSA · MANET · El-Gamal · Cryptography · Empirical

1 Introduction An algorithm is a set of instructions that are executed sequentially. The instructions are followed by a set of rules. Algorithms are used for solving complex problems step by step. An algorithm helps to find out the time complexity, and the space complexity of a program. In cryptographic algorithms, many computational operations are used In MANET cryptographic algorithms are used to find to secure the communication between two nodes. One need to find how much time is taken for encryption and decryption of message while moving in the network so that the one with less complexity can be used.

A. Kumari (B) · P. Pranav · S. Dutta Department of Computer Science and Engineering, Birla Institute of Technology, Mesra, Ranchi, India e-mail: [email protected] P. Pranav e-mail: [email protected] S. Dutta e-mail: [email protected] S. Chakraborty Department of Mathematics, Birla Institute of Technology, Mesra, Ranchi, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_9

111

112

A. Kumari et al.

It is extremely hard to develop a new design using composite cryptographic techniques without sound security analysis behind it, usually based on cryptographic reasoning. One way to achieve this is by employing cyber confidentiality. Mathematics and number theory are inseparably linked to cryptography. As a result, it’s impossible to come up with a new design that uses composite cryptographic approaches without first doing a thorough security study, which is usually based on cryptographic reasoning. Learning from others is one technique to achieve this goal. Analyzing current MANET/WSN security measures, as well as gaining a better understanding of the network to improve it. Learn how cryptographic approaches work in conjunction with MANETs to create a secure service. Network performance, scalability, storage, and synchronization should all be acceptable [1]. Cryptography is classified into two classes one is symmetric cryptography, and the other is asymmetric cryptography. Symmetric cryptography encompasses one key which is used for encryption as well as decryption. Asymmetric cryptography has two keys one is a public key another is private; one is used for encryption, and the other is used for decryption. Algorithm evaluation is an important part of the computational complexity idea, which offers theoretical estimation for the specified sources of an algorithm to clear up particular computational trouble. Analysis of algorithms is the evaluation of the amount of time and space required to execute it. An entire analysis of algorithms for running time of a set of rules includes the subsequent steps: • Implement the set of regulations. • Determine the time required for each primary operation. • Identify unknown quantities that may be used to explain the frequency of execution of the number one operation. • Analyzed the unknown portions, assuming the modeled input. • Calculate the complete strolling time by multiplying the time utilizing the frequency for every operation, then adding all the products.

2 Literature Review In [2] Author shows hybrid cryptographic algorithms for records protection in MANETs. They worked on two algorithms RSA and DSA algorithms. The generated trace documents and scripts the everyday overall performance of the proposed routing approach is evaluated and in comparison, with the traditional comfortable routing approach. The comparative typical overall performance of the proposed and traditional model is summarized. In their proposed technique the Energy is Low, Packet Delivery Ratio and Throughput are immoderate as have a study to Traditional technique. The awareness [3] at the displacement between nodes. The set of rules now not most effective requests the blanketed nodes as regularly as possible, but the connections with dominator nodes are also stronger and greater solid. Thus, in dynamic

Empirical and Statistical Comparison of RSA and El-Gamal in Terms …

113

MANET surroundings, the algorithm can better find the appropriate dominator node to assemble the CDS. However, the algorithm does not offer a development to the electric power levels offering increases in radio range and the quantity of the included nodes. It remains viable to boom the power consumption quickly, leading to extra radio noise, and the battery lifetime for cellular gadgets stay a vital yet unaccounted for factor. In [4] authors have presented a couple of constraint algorithms for multicast visitors engineering in MANET. The proposed algorithm is a new edition of more than one constraint QoS multicast routing optimization set of rules in MANET primarily based on GA (MQMGA). The proposed MQMGA optimizes the maximum hyperlink usage, the value of the multicast tree, the choice of the lengthy-life route, and the average delay and the most end-to-end-to-quiet put-off. The performance evaluation of our proposed techniques is completed thru modeling and simulation. The simulation effects exhibit that the proposed approach is an accurate and green technique for estimating and comparing the direction stability in dynamic mobile networks. In [5] author proposed work to decrease the chance of network degradation provoked by the strength exhaust of node. The use of the system learning algorithm allows us to determine the ultimate route in phrases of header range; specifically, they showed some properties for acquiring the optical course in phrases of the header quantity which refers to the mobility of the node. An approximate expression for the greatest sum of header variety is realized. Finally, based totally on our findings, they proposed a method to discover and select the course that accounts for the anticipated statistics transfer time over the direction permits enhancing the throughput, packet delivery ratio, and overhead of reactive routing protocols. In [6] delay optimization method in MANETs for multimedia transmission. The method maintains the sequence of in-order packets transmitted closer to the vacation spot which is received in a haphazard way to the buffer. Decrease the delay of packets with the aid of maximizing the in-order packets and minimizing the disorganized packets inside the buffer. A dynamic Knapsack set of rules is used to decrease packet loss due to the random motion of packets inside the buffer. Moreover, the method will increase the available space within the buffer by keeping a greater wide variety of inorder packets. Additionally, they proposed the mathematical relation of buffer size, packet size, and delay this is a critical metric for real-time multimedia applications. In [7] proposed a delay-sensitive segment scheduling algorithm (DSSSA) for presenting well-timed P2P streaming services in MANETs. To use the wireless useful resource more correctly, DSSSA adopts the approach of scheduling the segments flippantly transmitted into a MANET based totally on the playback price of the welltimed P2P streaming services. The method can use the limited bandwidth of MANET greater correctly. On the alternative hand, considering that DSSSA is performed irregularly and the facts the use of for the section schedule is maintained by using the history data received from the previous real section transmission, DSSSA is adaptive to host mobility and peer interaction. Statistical similarity among compounding procedures for RSA, a well-known cryptographic set of principles, is discussed in [8]. The two procedures used in encryption and decryption, namely electricity and modular mathematics, have been

114

A. Kumari et al.

analyzed, and the results show that they are statistically distinct. However, whereas the modular arithmetic operation is predicted to be an even more dominating of the two, this is not the case in practice. In addition, the fundamental theorem of the finite distinction method was applied to empirically assess the running time of both processes separately. The empirical run time for power operation is O (n), but the empirical run time for the mod operation is O (n) (1). In [9] comparing the overall performance of the LUC set of rules, Elgamal and RSA in the method of encrypting secret messages in textual content form have reached pretty a few conclusions. In phrases of the calculation manner, the LUC algorithm, Elgamal, and RSA have something in common, i.e. have a crucial generator within the structure of pinnacle numbers and the results of randomizing public and personal, and systematic keys. Based on the take a look, the quickest encryption system time is the RSA algorithm with 0.19 ms and 0.17 ms. In addition, the fastest decryption way is the LUC algorithm with 0.31 ms and 0.17 ms. In [10] RSA, single encryption algorithms are used. ECC may be carried out in different ways. ECC uses mathematics algorithms as the primary goal operations for excessive-degree safety functions consisting of encryption for gaining confidentiality and a virtual signature for authentication. ECC can be carried out in software and hardware. ECC follows a usual process where parties agree on publicly-recognized information items, and each consumer generates their public and private keys. In [11] paper offers relative analyses of AES, Twofish, RSA, and ElGamal cryptosystems. Predicated on the algorithms, new hybrid cryptosystems are Twofish RSA, AES RSA, and AES ElGamal. Memory consumption, translated report length, safety position, and encryption speed; those standards have been used to estimate the aboveproposed algorithms and hybrid fashions. After exploration, among the surpassed new hybrid models, AES RSA takes all blessings from the use of symmetric and asymmetric systems, so it is considered; however, Twofish RSA hybrid cryptosystem is quick. For destiny work, proposed hybrid fashions can be analyzed through the entropy index. With entropy, studies will be feasible to estimate the resistance of each set of rules against one-of-a-kind sorts of assaults, ordinarily in opposition to cipher text frequency evaluation. RSA [12] is a well-known public-key algorithm utilized by many applications. Security of RSA lies in the issue of factoring large numbers into prime factors. ElGamal Algorithm is a public key set of rules used for virtual signatures. Security of the ElGamal Algorithm lies in the difficulty of calculating the discrete logarithm. The proposed set of rules, a mixture of RSA and ElGamal Algorithm, has double protection; that’s the problem of factoring massive numbers into prime factors and calculating the discrete logarithm. Based on the work that has been performed, the aggregate of RSA and ElGamal for the important thing technology set of rules has decreased the computing time required. But, the security component is not pretty confirmed. It may be proved via doing some piercing exams to aspect r and calculate discrete logarithms. In this paper [13] they have virtualization of space in cloud computing for reduction of cost-efficient and making it more prominent for the research area. For securing the data in the cloud they first give the user authentication, they encrypt the stored

Empirical and Statistical Comparison of RSA and El-Gamal in Terms …

115

data with the help of the Elgamal encryption technique then they added the fuzzy rules-based integer for adding more security to the data. They have implemented their proposed work in python and evaluated it with the help of Matlab. The performance result is good in terms of various aspects like security, execution time, and cost association as compared to the existing methodology.

3 Run Time Comparison of RSA and EL Gamal Protocols Used in MANET The theoretical time complexity of RSA is based on message length and time taken for encryption and decryption of the message. For the encryption and decryption process, it will take different times so the complexity will also change. The time complexity for encryption will be calculated as C = Me mod n is O (log (N) 2 ) this is for fixed message length. Where e is the same order as N then the complexity of encryption will change to O (log (N) 3 ). For the decryption of the message, the complexity will change. M = Cd mod n this formula is used for decryption of the message. The time complexity is O (log (N)3 ). El-Gamal algorithms are based on asymmetric cryptography in which two keys are used for encryption and decryption. The time complexity of El-Gamal is Big O which is written as O(n). The time complexity of algorithms depends on many factors like the length of the message, hardware of the system, and operating system. While algorithms are executing in the system it depends upon the number of operations it is performing. Length of the message increases then the execution time will increase.

4 Fundamental Theorem of Finite Difference Let f be a real-valued function and let a and b be integers such that a ≤ b. If F =  − if then



bn = af(n) = F(b + 1) − F(a).

The fundamental theorem says that if the nth variation of an nth degree polynomial is constant and the higher differences are 0 then, n Pn (x) = C = Constant n+r Pn (x) = 0, r = 1, 2, 3 . . . Here  is the forward difference operator, state that  f(x) = f(x + h) − f(x) for the values of f(x) calculated as x = a, a + h, a + 2 h … h being the interval of differencing.

116

A. Kumari et al.

We analyzed the time complexity of RSA and El Gamal empirically using the fundamental theorem of finite difference. We run both algorithms 3 times and took the mean of the 3 trails as the final execution time. Then we calculated the first and the second difference in the mean execution time.

5 Result and Analysis In this encryption and decryption table of RSA as in Tables 1 and 2, we have taken three trails for different message lengths. The time for different lengths of the message is different. After taking the number of trials time varies for encryption and decryption of the message. Then we have taken the mean of three different trails. Then take the first difference of mean was and then take the second difference. The result of the second difference is almost constant which shows Big O which means the worst case of the algorithm. The time complexity of RSA for encryption is Oemp (n2 ). Table 1 Encryption time for RSA Encryption time RSA Message bit

Trial 1

Trial 2

Trial 3

Mean

8 16

First difference ( x)

Second difference ( x2 )

0.000330

0.00050

0.00050

0.00044

0.00012

0.00032

0.000931

0.00040

0.00036

0.00056

0.00044

− 0.00025

32

0.000486

0.00130

0.00123

0.00101

0.00019

0.00036

64

0.000799

0.00095

0.00184

0.00120

0.00055

0.00229

128

0.002654

0.00141

0.00115

0.00174

0.00283

0.00918

256

0.000885

0.01129

0.00155

0.00457

0.01201

− 0.01201

512

0.003270

0.00321

0.04328

0.01659

Table 2 Decryption time for RSA Decryption time RSA Message bit Trial 1

Trial 2

Trial 3

Mean

First difference ( Second difference x) ( x2 )

8

0.00075 0.00039 0.00042 0.00052 0.00002

0.00074

16

0.00121 0.00022 0.00019 0.00054 0.00076

− 0.00060

32

0.00028 0.00184 0.00178 0.00130 0.00016

0.00007

64

0.00049 0.00039 0.00348 0.00145 0.00023

− 0.00013

128

0.00183 0.00175 0.00145 0.00168 0.00010

− 0.00002

256

0.00140 0.00209 0.00185 0.00178 0.00008

− 0.00008

512

0.00173 0.00143 0.00242 0.00186

Empirical and Statistical Comparison of RSA and El-Gamal in Terms …

117

Table 3 Encryption time for EL GAMAI Encryption time EL GAMAI message bit

Trial 1

Trial 2

Trial 3

Mean

First difference ( x)

Second difference ( x2 )

8

0.009293

0.007745

0.006064

0.0077007

0.00563

16

0.014014

0.015139

0.010836

0.0133297

− 0.00032

0.00530733

32

0.014075

0.012214

0.012748

0.0130123

0.00499

− 0.00330100

64

0.017437

0.017086

0.019484

0.0180023

0.00169

0.00151067

128

0.019169

0.019341

0.020564

0.0196913

0.00320

− 0.00024600

256

0.023099

0.023005

0.022569

0.0228910

0.00295

− 0.00295367

512

0.025083

0.026891

0.02556

0.0258447

− 0.00594633

Table 4 Decryption time for EL GAMAI Decryption time EL GAMAI Message bit Trial 1

Trial 2

Trial 3

Mean

First difference Second ( x) difference ( × 2) − 0.00267

8

0.004315 0.003949 0.005093 0.004452 0.00294

16

0.007807 0.008239 0.006132 0.007393 0.00027

32

0.00791

0.007559 0.007518 0.007662 0.00315

− 0.00202

64

0.010452 0.010262 0.011724 0.010813 0.00113

0.00095

128

0.01231

0.011792 0.011947 0.00208

− 0.00008

256

0.015112 0.013585 0.013387 0.014028 0.00200

− 0.00200

512

0.015088 0.017087 0.015912 0.016029

0.01174

0.00288

In El-Gamal algorithms as in Tables 3 and 4, we have taken three different trails for different message lengths for encryption and decryption. The Time taken for encryption and decryption of messages is different for different message lengths. After taken of three trials then taking the mean of three trials and then calculated the first difference and then taken second difference. The time variation is almost constant which shows that El-Gamal encryption and decryption time complexity is Big O which shows the worst case of the algorithm. The time complexity of the algorithm is Oemp (1). The validated results are shown in the Figs. 1, 2 3 and 4.

6 Conclusions and Future Work This paper analyzed that theoretical time complexity is different from real-time complexity. RSA and El-Gamal both algorithms are asymmetric cryptography.

118

Fig. 1 Fitted line plot of encryption execution time of RSA

Fig. 2 Fitted line plot of decryption time of RSA

A. Kumari et al.

Empirical and Statistical Comparison of RSA and El-Gamal in Terms …

Fig. 3 Fitted line plot of encryption time is non-linear of EL-GAMAL

Fig. 4 Fitted line plot of decryption time of EL-GAMAL

119

120

A. Kumari et al.

Encryption and decryption time is different for different algorithms. The theoretical time complexity of RSA is O(n3 ) and for El-Gamal is O(n) where n varies. According to the practical implementation of RSA and El-Gamal, the time complexity is different. By using the fundamental theorem of the finite difference method, we analyzed that time complexity is different. The empirical time complexity of RSA is Oemp (n2 ) and for El-Gamal is Oemp (n2 ). According to empirical analysis, the time complexity is different from theoretical.

References 1. Jianmin C, Wu J (2005) A survey on cryptography applied to secure mobile ad-hoc networks and wireless sensor networks. In: Proceedings of CRYPTO, pp 27–56 2. Sharma A, Bhuriya D, Singh U (2015) Secure data transmission on MANET by hybrid cryptography technique. In 2015 International conference on computer, communication, and control (IC4), pp 1–6 3. Leu S, Chang RS (2012) A weight-value algorithm for finding connected dominating sets in a MANET. J Netw Comput Appl 35(5):1615–1619 4. Sun B, Pi S, Gui C, Zeng Y, Yan B, Wang W, Qin Q (2008) Multiple constraints QoS multicast routing optimization algorithm in MANET based on GA. Prog Nat Sci 18(3):331–336 5. Arora SK, Monga H (2016) Performance evaluation of MANET based on knowledge base algorithm. Optik 127(18):7283–7291 6. Ahmad SJ, Reddy VSK, Damodaram A, Krishna PR (2015) Delay optimization using the Knapsack algorithm for multimedia traffic over MANETs. Expert Syst Appl 42(20):6819–6827 7. Hu CC, Lai CF, Hou JG, Huang YM (2017) Timely scheduling algorithm for P2P streaming over MANETs. Comput Netw 127:56–67 8. Pranav P, Dutta S, Chakraborty S (2021) Empirical and statistical comparison of intermediate steps of AES-128 and RSA in terms of time consumption 9. Sari PP, Nababan EB, Zarlis M (2020) Comparative study of Luc, ElGamal, and RSA algorithms in encoding texts. In: 2020 3rd international conference on mechanical, electronics, computer, and industrial technology (MECnIT). IEEE, pp. 148–151 10. Mallouli F, Hellal A, Saeed NS, Alzahrani FA (2019) A survey on cryptography: a comparative study between RSA vs ECC algorithms, and RSA vs El-Gamal algorithms. In: 2019 6th IEEE International conference on cyber security and cloud computing (CSCloud)/2019 5th IEEE international conference on edge computing and scalable cloud (EdgeCom). IEEE, pp 173–176 11. Jintcharadze E, Iavich M (2020) Hybrid implementation of Twofish, AES, ElGamal, and RSA cryptosystems. In: 2020 IEEE East-West design & test symposium (EWDTS). IEEE, pp 1–5 12. Iswari NMS (2016) Key generation algorithm design combination of RSA and ElGamal algorithm. In: 2016 8th International conference on information technology and electrical engineering (ICITEE). IEEE, pp 1–5 13. Pandian AP (2020) Development of secure cloud-based storage using the Elgamal hyper elliptic curve cryptography with fuzzy logic-based integer selection. J Soft Comput Paradigm 2(1):24– 35

IoT Communication to Capture and Store Data to Thingspeak Cloud Using NodeMCU and Ultrasonic Sensor Priya J. Payyappilly and Shweta Dour

Abstract Internet of Things is a domain which has gained quite a lot of momentum in the past decade. The ability of things communicaiting via a network or the internet is both a fascinating and a challenging concept. The authors in this study aim to introduce users to the concept of Internet of Things and demonstrate a use-case wherein data can be uploaded to the cloud; specifically, here the Thingspeak cloud is used as a storage for the data that is uploaded from an ultrasonic sensor via a microcontroller NodeMCU and WiFi standard. This usecase will help new users understand the scope and capability of communication in the Internet of Things. The exeprimental setup shows successful communication between nodeMCU and thingspeak cloud through the internet. Keywords Internet · Internet of things · Thingspeak · Communication · Cloud · NodeMCU · Sensor · Ultrasonic sensor

1 Introduction The Internet revolution [1] in the early nineties brought about a new approach for remote sharing and storing of data. A person practically in any part of the world could get access to this data via a worldwide network of interconnected computers forming a kind of sophisticated information superhighway. With the advancement and subsequent amalgamation of different technologies like integrated circuit technology, mobile technology and computing power, the Internet has developed magnificently in a short span of four decades and has become a part of routine life like a utility. The Internet initially was conceptualized to facilitate sharing of information by interaction between machines. This concept was further expanded by Ashton [2], P. J. Payyappilly (B) · S. Dour Navrachana University, Vadodara, Gujarat, India e-mail: [email protected] S. Dour e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_10

121

122

P. J. Payyappilly and S. Dour

in the late nineties, to include things into the Internet domain. He envisioned a world where physical and virtual ‘things’ [2] could be connected via Radio Frequency Identification (RFID) technology and make decisions on their own to accomplish certain tasks. He coined the term Internet of Things [2], abbreviated as IoT to distinguish this technology from traditional networking and internet. Since then, this concept has been taken up by various organizations for implementation and standardization. However, a shift from conventional networking and internet to IoT requires lot more efforts involving combining of a wide range of devices and technologies. It can be stated that IoT is still in an evolutionary stage, involving myriad levels of complexities and challenges. The authors in this article discuss the concept of IoT in depth and highlight the main aspects of this study. On generalizing from the perspective of technology, IoT is a combination of the elements as shown in Fig. 1. It consists of hardware, software, network, cloud and analytics. • Hardware—Hardware devices used for IoT implementation include microcontrollers and microprocessors like Arduino, Beagle Bone and Raspberry Pi. Microcontrollers and microprocessors act as processing elements of the system. These devices perform computation and execute programs to achieve some goal. These devices are interfaced with sensors and actuators. Sensors are instruments that measure certain physical quantities and represent them in digital form. Actuators are devices that manoeuver on getting a signal. The devices mentioned above can

Fig. 1 IoT, a technological overview

IoT Communication to Capture and Store Data to Thingspeak Cloud …

• •

• •

123

be embedded into a larger system or can function individually depending on the application that they are used in. Sensors and actuators work in tandem according to the signal from processing elements. Software—It’s the soul of the system, software encompasses the operating system, programs for performing operations according to input, the commands and also may include APIs for interfacing various technologies. Network—Network is the collection of technologies and protocols that facilitate connection between different devices. It includes wired and wireless connecting elements, protocols for enabling the connections like MQTT, HTTP, routing protocols, etc. It also includes the server which is a device for housing files. Cloud—Cloud is a collection of servers and networking resources offered to users on a pay as you use basis to enable ease of storage and access of data. Analytics—It is the scientific method of analyzing data that is either in transit on the network or stored on the cloud. It uses various algorithms entailing machine learning and deep learning to generate insights from available data. It is important to note here that all of these technologies cannot exist as solitary, there is a great degree of overlap among all of them. Also, there are other protocols and technologies that are used in assistance to the ones mentioned above but here, we are including only the primary technologies. Figure 1 is designed to show its general idea and simplicity from a technological perspective. IoT can also be depicted in the form of a layered approach. Layered approach is a way of representing any technology by dividing it into layers. Layers indicate demarcation points as well as modules of a large technology. Each layer performs a dedicated task and provides a set of services to its adjacent layers or higher layers. Generally, there are three major layers as shown in Fig. 2, the device layer, the network layer and the cloud. Each of these layers have their associated sub layers and perform functions that are explained below:

Fig. 2 Layered architecture of IoT

124

P. J. Payyappilly and S. Dour

Device Layer: Consists of the devices that make up the implementation of IoT. This layer generally has sensors, actuators or a combination of both in the form of embedded systems and most importantly controllers. The term controllers is a broad classification. Controller can be broadly include microcontrollers and microprocessors. Network Layer: Consist of the physical network between devices that are set up using various networking standards and protocols. It also consists of networking devices like routers and switches which facilitate the transmission of packets. Cloud: The cloud is the term used to describe cloud computing. Cloud computing is the on-demand delivery of compute power, database, storage and infrastruce. It is a service that is provisioned by various companies on a pay by use policy. The users only pay for the resources they use. This practice greatly diminishes the requirement of configuring dedicated hardware and resources; saving money in the process. Cloud in IoT is used for one additional function apart from storing the data, it also analyzes the data to get useful insights. These insights play a major role in organizational decision making as the IoT solution will always be implemented by an organization seeking to service customers. As IoT is a confluence of various technologies, the authors through this study aim to: Introduce the readers to the concept of Internet of Things, explain the layered architecture of IoT, present the design and working of NodeMCU microcontroller and an ultrasonic sensor, familiarize the readers to the Thingspeak cloud platform and finally deploy an application to store the data obtained from the ultrasonic sensor to thingspeak via the internet.

2 Related Work Various attempts at studying as well as implementing the idea of IoT using the concepts stated in the aforementioned sections are in the budding stage [3]. The primary focus of this literature survey by the authors is to explore different projects that are implemented using NodeMCU and Thingspeak. To summarize the contributions of various researchers in this domain, Ramdevi et al. [4] have implemented an indoor air quality and noise monitor using NodeMCU and Blynk application. NodeMCU is interfaced with DHT 11 which is a temperature and humidity sensor [5]. In addition to this, NodeMCU is also interfaced with MQ-3 sensor which is a gas sensor and SEN-12642, which is noise sensor. The cloud platform used to store data and obtain its visualization is the blynk platform [4]. Using the sensors and microcontroller, the system is able to sucessfully produce an outcome in the form of an alert having pollution and noise level values. Additionally, they add a buzzer as an alarm to notify the users apart from e-mail and message notifications. Juan et al. [6] develop a modular and economical energy management system for ameliorating the electricity usage caused by electric water heaters. The authors state that energy consumption due to electric water heaters is around 30c/o of the total energy consumption; therefore, it is imperative that some measures have to be taken to curb

IoT Communication to Capture and Store Data to Thingspeak Cloud …

125

the energy usage so that it is lighter on the expenses. The complete system consists of NodeMCU microcontrollers connected to various electronic appliances and communicating with Raspberry Pi to provide status and energy usage of those appliances. The results show promising projections for annual savings using this energy management system. Ramaiah et al. [7] in their implementation have designed a tracking system for patient medication using NodeMCU interfaced with temperature sensor, RTC, LCD and servo motor. It notifies the patient at the time of taking their medicine. The box has a buzzer attached to it that persists ringing until the box is opened and it doesn’t allow the box to be opened on a time other than the alloted time for medicine consumption. Jayaysingh et al. work along similar lines in their implementation of patient monitoring system [8] which monitors the vital signs of a patient using NodeMCU, pulse rate sensor, the cloud and smartphone. This is a very useful application of IoT which can redefine the healthcare realm and patient care solutions. Kharade et al. [9] implement a prototype for checking fire safety and air quality. The authors state that the rise of industrialized enviornment and emission of gases from different mechanized devices taint the atmosphere, so they execute a safety mechanism which has nodeMCU interfaced with flame sensor, DHT11 sensor and air quality sensor. These are interconnected to a cloud to complete the prototype of an IoT system. The cloudplatforms used here are blynk and Thingspeak. Blynk is used to alert the user while Thingspeak is used to store the data. Lufyagila et al. [10] use nodeMCU interfaced with various sensors effectively to monitor enviornmental conditions in poultry farms. The case study is specifically related to Tanzania, Africa, in which sample data has been taken from the farmers of Kilimanjaro and Arusha region. Conventional method of maintaining poultry is replaced by cost-effective IoT solution of enviornmental condition monitor. It has a NodeMCU connected with various electrical devices using relay. The real-time statistics are sent to a database server and raspberry Pi via a wireless access point. Raspberry Pi here is the gateway server which communicates to the end devices. The data is periodically synchronized to the end devices. If there is no connection to the internet, the data is stored on the local gateway and synchorized on availability of internet. The available information can then be visualized on the end devices for necessary action on the user’s part. The results of this implementation show that the proposed system saves 84% time and saves 66.7% labour costs. Kaushik et al. [11] have implemented an IoT based system to monitor the power consumption in buildings; enabling the concept of smart buildings. The IoT solution detects human presence in a room using thermal sensor and Machine learning technique and SVM. Based on the detection, it switches the light on or off. It uses raspberry Pi and nodeMCU to compute and perform the necessary action with or without human presence. From the logged data, the reason for power consumption in a building can be known. The aforementioned realization of IoT solution helps in mitigating the useless expending of energy. Gupta et al. [12] have developed a testing system for monitoring and acquiring data out of a Photovoltaic cell in harsh enviornmental conditions. They have used Photovoltaic cell that is connected to four NodeMCU boards interfaced with different sensors like DHT22, dust sensor, wind sensor, pyranometer and INA 219 sensor. The software part of these devices is connected to the Thingspeak cloud and blynk application via

126

P. J. Payyappilly and S. Dour

the internet. There is a WiFi switch for enabling and disabling WiFi. This feature is added so that energy can be saved. The experimental results show that the measurements obtained by the aforementioned system is better than traditional methods and is resilient in challenging enviornmental conditions. A part of home automation system is presented by Garg et al. [13]. The authors propose a system that has a NodeMCU connected to different sensors and continuously monitors surrounding parameters like gas, temperature, humidity and light. Through IFTTT and Blynk application. Through the use of this system, real-time data generated at home can be monitored remotely from anywhere by internet connectivity. Vijaylakshmi et al. [14] in their work have developed an intelligent system to study different parameters that affect ease of learning in a classroom enviornment. Arduino microcontroller is interfaced with different sensors which measure sound levels, temperature, etc., in a room and show visualization on Thingspeak platform about the different parameters. The results are then displayed on LCD monitor. The visualization is helpful in getting to know the enviornment and help in increasing the comfort in the enviornment. Amala et al. [15] have devised a technology for continuously checking soil moisture content using Arduino which is then published to Thingspeak. They have classified the whole system into four layers, namely: sensor layer, Middleware, Communications layer and Cloud and application layer. The comparative analysis of each of the work presented above has been highlighted in Table 1. As shown in Table 1, it can be inferred that a very few projects have used combination of NodeMCU and Thingspeak. Also the use of ultrasonic sensor has not been seen in amy of the aforementioned articles. In line with the existing literature, the Table 1 Comparative analysis of communication methodologies in the literature survey Article

Project

Technology

[4]

Indoor air quality and noise monitor

nodeMCU and Blynk, DHT11, MQ3

[6]

Electric energy tracker

NodeMCU and Raspberry Pi

[7]

Patient’s medicine tracking system

NodeMCU, RTC, LCD, servo motor

[8]

Patient’s vital signs monitor

NodeMCU, cloud, pulse rate sensor and smartphone

[9]

Air quality monitoring

NodeMCU, blynk and Thingspeak

[10]

Enviornmental conditions monitoring in poultry farms

NodeMCU, Raspberry Pi

[11]

Smart Buildings

NodeMCU, Raspberry Pi, thermal sensor

[12]

Procuring data from a photovoltaic cell

NodeMCU, Thingspeak, Blynk, photovoltaic cell

[13]

Home automation

NodeMCU, Blynk, IFTTT, gas, temperature, humidity and light sensors

[14]

Learning comfort in classroom

Arduino, Thingspeak

[15]

Soil moisture sensing

NodeMCU and Thingspeak, moisture sensor

IoT Communication to Capture and Store Data to Thingspeak Cloud …

127

present article focuses on establishing connection between the cloud and a device. The device here is a combination of microcontroller NodeMCU and ultrasonic sensor.

3 Proposed Work The setup consists of NodeMCU interfaced with ultrasonic sensor which in turn is connected to the Thingspeak cloud via the internet. If we try to explain it in terms of the layered architecture of IoT as shown in Fig. 6. The mapping of various layers for this experimental setup can be according to the one shown in Fig. 3. The device layer is the lowest layer which consist of the apparatus making up the foundation of IoT technology. It consists primarily of the microcontroller NodeMCU and ultrasonic sensor. NodeMCU is an open source platform that is executed in conjunction with the ESP8266 WiFi SoC [16]. It can be programmed using the Lua scripting language although, the programming of it using Arduino IDE is widespread due to the developer communities efforts. An ultrasonic sensor which has two pins namely trigger pin and echo pin is also used. Ultrasonic sensor calculates distance of an object based on the reflected signal that is sent. This makes up the hardware part of the system. NodeMCU and ultrasonic sensor are as shown in Figs. 4 and 5. NodeMCU has been chosen as it is compatible, easy to interface and economical for prototyping. The network layer consists of the communication technologies which broadly speaking is the internet and WiFi. Thingspeak is used as the platform for cloud. It stores the data in a unit called channel. A channel is the basic building block of Fig. 3 Layered architecture of IoT mapping with experimental setup

128

P. J. Payyappilly and S. Dour

Fig. 4 NodeMCU ESP8266

Fig. 5 Ultrasonic sensor

Thingspeak. It consists of various attributes which a user can update to store relevant data. Initial Configuration of Thingspeak Channel Example of a Thingspeak channel is shown in Fig. 6 and 7. The attributes that are to be updated need to be activated first. Then they can be named and values for them can be entered. In the attributes of channel contain field numbers ranging from one to eight. The field is analogous to variable names in a program. As variables are named location of memory, similarly, field name is the named tag given for attribute storage. User can create upto 8 fields. Before naming a field, the user is supposed to activate it by selecting checkbox on the right side of the field. Only the activated field will store the value for any attribute.

IoT Communication to Capture and Store Data to Thingspeak Cloud …

129

Fig. 6 Thingspeak ınitial configuration: attributes for code

Initially, all attributes for storing data are not selected in the thingspeak channel so they can be seen as inactivated. For the proposed method, since only the distance has to be stored, a single field named distance has been created as shown in Fig. 8. The checkbox on the right has to be selected, at that time, the attribute will be activated. On entering the values and saving, the channel statistics is as shown in Fig. 8 is displayed. It contains channel ID which is a unique identifier of the channel. The name field is the identifier of the channel from the user perspective. Description field is analogus to comments in C program. The programmer can add a description to depict information about the channel for understanding what the use of the channel is. After that comes field 1 which is the variable holding value of distance of object from ultrasonic sensor. Since this is a simple experiment, a single field is only used to represent value of distance. Figure 9 shows the experimental setup where nodeMCU is interfaced with ultrasonic sensor. The nodeMCU and ultrasonic sensors are in their default configuration and no changes are done to their configuration. On connection with the computer system, the presets are selected automatically when nodeMCU board is selected in the Arduino IDE. The ultrasonic sensor has been selected because of its ease of interfacing with the NodeMCU board. The rationale for choosing NodeMCU for controller is that it is economical to purchase, is open source and can be used in conjunction with Arduino IDE easily.

130

P. J. Payyappilly and S. Dour

Fig. 7 Thingspeak ınitial configuration: other fields

The code of the proposed setup can be summarized through the flowchart shown in Fig. 10. The variables and hardware are initialized in the begining. Then an attempt to connect to WiFi is done. If it is sucessful, the loop function in Arduino IDE (Integrated Development Enviornment) executes continuously until there is some error or the connection no longer exists. The output is continuously printed on the serial monitor and is uploaded to Thingspeak every 15 s. The system continuously checks for broken communication or un-initialized variables. If any of the conditions are not fulfilled, exit message is printed on the serial monitor and the process is ended there. It is shown by terminator labelled End in the flowchart.

IoT Communication to Capture and Store Data to Thingspeak Cloud …

131

Fig. 8 Thingspeak configuration for IoT communication

Fig. 9 Experimental setup

The implementation in the form of code for this application is shown in Figs. 11 and 12. The system and its components are initialized by declaring variables and pins as shown in Fig. 11. Initially, the header files are included, then the constants are declared in the global declaration section. The calculation parameters and its formulae are declared thereafter. The rest of the code for communication with Thingspeak as well as displaying the values on serial monitor is given in Fig. 12. There are two main functions in Arduino program viz setup and loop. Setup function consists of all the functions that facilitate setup of a system. Here, as shown in Fig. 12, configuration with serial monitor, Thingspeak cloud, WiFi connection and configuration of pins on nodeMCU is done. Once these are successful, the loop function is used to continuously read values from ultrasonic sensor and upload to cloud. The aforementioned points are demonstrated in Figs. 11 and 12.

132

P. J. Payyappilly and S. Dour

Fig. 10 Flowchart for the proposed method implementation

Fig. 11 Code for IoT arduino IDE communication in with thingspeak: ınitialization

IoT Communication to Capture and Store Data to Thingspeak Cloud …

133

Fig. 12 Code for IoT arduino IDE communication in with thingspeak: setup and loop function

4 Results and Discussion Figures 13 and 14 show the results on execution of the program. Figure 13 shows the serial monitor in Arduino on which the values of distance of an object from centimeter to inches is displayed. The visualization shown in Fig. 14 is the line graph of the values that are obtained while the sensor is active and communicating to the microcontroller. Suitability of the proposed model in experiment with other sensors and application wise necessity The proposed model is suitable to be implemented with most of the other sensors like DHT sensor, gas sensor, touch sensor etc.; since nodeMCU is an open source technology, it can be easily interfaced with other sensors. If the other sensors are from a different vendor, the related header file for that sensor is to be downloaded in the Arduino IDE and it code can be easily uploaded without much hassle. For instance, if nodeMCU is interfaced with DHT sensor, only DHT header library has to be downloaded and included in the current program. Once this process is done, the program executes smoothly.

134

P. J. Payyappilly and S. Dour

Fig. 13 Output on arduino monitor

Fig. 14 Visualization on thingspeak

Ultrasonic sensors can be used in various applications like object detection, distance measurement and liquid level sensing, the application suggested in the present article can be used in conjunction with the aforementioned uses to yield better results and insights for the data that is produced.

5 Conclusion and Future Scope Internet revolution has unfolded new avenues for device communication on the network. This study attempts to familiarize the reader with the concept of internet of things and elicit an example of the same. The example is shown in terms of communication between an ultrasonic sensor, nodeMCU and Thingspeak cloud. The experimental setup and results show that data is sucessfully uploaded to the cloud using API keys avalable on Thingspeak channel. However, the delay in uploading

IoT Communication to Capture and Store Data to Thingspeak Cloud …

135

of data on the cloud platform can be a challenge when real time data has to be given to the cloud and if any decisions have to be made using that real-time data. Also a limitation is security of the devices as well as the communication process. There is ample scope for exploration in the area of security of devices and communication. Futhermore, this study can be extended to include notification through IFTTT if any event occurs and also generating insights from the uploaded data through MATLAB on thingspeak. The study can also incorporate different sensors along with nodeMCU in different applications to upload data to Thingspeak.

References 1. Kleinrock L (2010) An early history of the internet [History of Communications]. IEEE Commun Mag 48(8):26–36 2. Goyal P, Sahoo AK, Sharma TK (2021) Internet of things: architecture and enabling technologies. Mater Today: Proc 34:719–735 3. Smys S, Raj JS (2019) Internet of things and big data analytics for health care with cloud computing. J Inf Technol 1(1):9–18 4. Ramdevi M, Gujjula R, Ranjith M, Sneha S (2021) IoT evaluating ındoor environmental quality check of air and noise. Mater Today: Proc. ISSN 2214-7853 5. Duraipandian M, Vinothkanna R (2019) Cloud based internet of things for smart connected objects. J ISMAC 1(02):111–119 6. Tejero-Gómez JA, Bayod-Rújula AA (2021) Energy management system design oriented for energy cost optimization in electric water heaters. Energ Buildings 243(111012). ISSN 03787788 7. Challa R, Yamparala R, Kanumallı SS, Kumar KS (2020) Advanced patient’s medication monitoring system with ardunio UNO and NODEMCU. In: 2020 4th International conference on electronics, communication and aerospace technology (ICECA), pp 942–945. https://doi. org/10.1109/ICECA49313.2020.9297420 8. Jayaysingh R, David J, Raaj MJM, Daniel D, BlessyTelagathoti D (2020) IoT based patient monitoring system using nodeMCU. In: 2020 5th International conference on devices, circuits and systems (ICDCS), pp 240–243. https://doi.org/10.1109/ICDCS48716.2020.243588 9. Kharade M, Katangle S, Kale GM, Deosarkar SB, Nalbalwar SL (2020) A nodeMCU based fire safety and air quality monitoring device. In: 2020 International conference for emerging technology (INCET) 2020, pp 1-4. https://doi.org/10.1109/INCET49848.2020.9153983 10. Lufyagila B, Machuve D, Clemen T (2021) IoT-powered system for environmental conditions monitoring in poultry house: a case of Tanzania. Afr J Sci Technol Innov Dev. https://doi.org/ 10.1080/20421338.2021.1924348 11. Kaushik S, Srinivasan K, Sharmila B, Devasena D, Suresh M, Hitesh Panchal R, Ashokkumar KK, Sadasivuni & Neel Srimali, (2021) Continuous monitoring of power consumption in urban buildings based on internet of things. Int J Ambient Energy. https://doi.org/10.1080/01430750. 2021.1931961 12. Gupta V, Sharma M, Pachauri RK, Dinesh Babu KN (2021) A low-cost real-time IoT enabled data acquisition system for monitoring of PV system. Energy Sources Part A: Recovery Utilization Environ Eff 43(20):2529–2543. https://doi.org/10.1080/15567036.2020.1844351 13. Garg S, Yadav A, Jamloki S, Sadana A, Tharani K (2020) IoT based home automation. J Inf Optim Sci 41(1):261–271. https://doi.org/10.1080/02522667.2020.1721581 14. Vijayalakshmi R, Jayasimman L (2021) The design of IoT based smart learning environment for learner’s comfort level monitoring. Mater Today: Proc. ISSN 2214-7853

136

P. J. Payyappilly and S. Dour

15. Praveen AAA, Tamilnesan P, Muthukumaran M, Udayakumar MD (2021) Experimental analysis of moisture content with involuntary irrigation structure in soil, Mater Today: Proc 45, Part 2, 1893–1897. ISSN 2214-7853 16. Nooruddin S, Islam MM, Sharna FA (2020) An IoT based device-type invariant fall detection system, Internet of Things 9(100130). ISSN 2542-6605

A Comprehensive Study on Cloud Computing: Architecture, Load Balancing, Task Scheduling and Meta-Heuristic Optimization Shruti Tiwari and Chinmay Bhatt

Abstract Cloud computing (CC) is evolving computing model with a vast array of heterogeneous autonomous systems by modular computational architecture. Load balancing of activities on the cloud environment is an essential part of distributing services from the data center. CC is agonized by overloading demands because of dynamic computing through the internet. Load balancing must be done to ensure maximum use of the resources in all virtual machines (VM). Task scheduling is a crucial step for improving cloud computing’s overall efficacy. Task scheduling is therefore significant to minimize energy usage and increase service providers’ benefit by reducing the time required. This work provides a detailed study about the cloud computing architecture, load balancing (LB) mechanism, task scheduling (TS) framework in the cloud environment. Various meta-heuristic optimization techniques have been implemented to manage the load over virtual machines using task scheduling and load balancing terminologies. Various research gaps and issues have been identified from the literary work done by various researchers. This comprehensive study has motivated and provided us future direction to do work in this field. Keywords Cloud computing · Virtual machines · Task scheduling · Load balancing · Meta-heuristic optimization

1 Introduction Cloud Computing [1, 2] is an evolving software deployment and maintenance trend that is being embraced by industries like Microsoft, IBM, Google, and eBay. IBM-Blue Cloud framework, eBay Cloud, Google App Engine, as well as Distributed Computing Platform are many conceptual systems and frameworks. S. Tiwari (B) · C. Bhatt Department of Computer Science Engineering, RKDF College, Bhopal, (M.P.), India e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_11

137

138

S. Tiwari and C. Bhatt

Cloud Computing is seen as the upcoming development which would influence organizational organizations and also how they handle their IT infrastructure and services. A key researching area is infrastructure and technology which cloud services and deployed models have presented. The Internet has always been a driving factor in the advancement of different technologies. Cloud Computing is arguably the most debated amongst all these. The cloud computing model has seen an immense shift toward this usage over the last several months and has become a standard in the IT space as it offers its users and suppliers substantial cost savings including new business potential [3]. The benefits that use cloud computing are including: (1) Minimization in the maintenance cost and its hardware cost. (2) Accessible to all. (3) Flexible including automatized procedures in which the client does not have to think about complex problems such as software up-gradation [4, 5]. Although the concept of cloud computing has various variations, this new computing model is defined by certain fundamental concepts. Cloud Computing offers technical resources that are provided on-demand as-a-service through the use of the Web, typically managed across premises. Big Evolution in the field of computer science, a service. CC can efficiently and securely deliver various IT services and resources according to the requirement of the user. Cloud computing offers usagedriven applications. It provides services like Infrastructure as a Services (IaaS), Platform as a Services (PaaS), and Software as a Services (SaaS) [6]. Clients of such resources don’t own infrastructure in the remote cloud, yet pay for resources on the per usage, provided that a 3rd party operates and maintains the public cloud. The main principle, therefore, is the virtualization of resources. They rent the physical infrastructure, systems, and software inside a shared structure in a realistic situation. From virtual networks, computing systems, integrated data center, end-user web apps, and web services to immense computing-oriented service, security features will differ. In many fields of ITs, cloud computing can be used to resolve issues such as GIS, Science Research, Decision Making, ERP, e-Governance Systems [7], Web App Creation, Mobile Technologies, etcetera. As the main back-end computing infrastructure, cloud computing depends on data centers [8]. With the need for cloud computing more dramatically in recent years, geographically dispersed data centers are offering more and more cloud resources to improve the reliability and consistency of services. A huge amount of energy is essential to operate these geographically dispersed data centers, which represent around 15% of the overall cost of a data center (DC) [9]. However, considering data centers as supercomputers with shared resources or making the second stage simple, consistent with servers, new approaches to geographical load-balancing (GLB) as in Fig. 1 [10] mostly concentrate on the first phase, i.e. the proportion of work demands that are assigned to each data center, of employment planning. Moreover, the existing GLB approaches often rely on a specific resource dimension (for example, CPU) and consistent resource need,

A Comprehensive Study on Cloud Computing: Architecture, Load …

139

Fig. 1 Cloud load balancing

without taking into account the requirements of many separate jobs and heterogeneous resources. This is significant, in particular, because modern data centers are usually built from a variety of server groups with various processing capacity, size of memory, and storage spaces specifications [11]. The diversity of the market profile for workloads for different occupations exacerbates such heterogeneity further, e.g. CPU intensive numerical computing activities are generally required while high memory support is normally required for database operations [12]. Load must be spread among the cloud-based networks such that there must be no over-utilized or underutilized nodes within the networking process. LB is a common cloud problem that makes it difficult for applications adjacent to QoS (quality of service) measurement to sustain performance according to the SLA (service level agreement) document essential by enterprise cloud providers. Cloud providers have difficulty distributing similar workloads through servers. Effective LB technology can maximize and guarantee high user satisfaction by effectively using VMs resources [13]. In the cloud environment, a load must be balanced through various techniques. Algorithms of LB may be split into Algorithms for dynamic LB and static LB. Load is spread by previous knowledge and information in the Static LB Algorithm. Static algorithms at the same time do not take account of the current workload [14]. These algorithms are for the less work-loaded cloud. Dynamic LB Algorithms take into account current work Cloud load. Centralized and Semi-distributed LB algorithms can be categorized into dynamic LB algorithms [15]. Intelligent strategies like GA (Genetic Algorithms), PSO (Particle Swarm Optimization), ACS (Ant Colony System), and ABC (Artificial Bee Colony) can address problems of load balancing [16]. Various analysis was undertaken in the area of LB and TS in the cloud environments.

140

S. Tiwari and C. Bhatt

The task scheduling [17], known for providing essential cloud service efficiency, currently represents a related hot topic. However, due to inappropriate scheduling, the dilemmas of resources that are underused (unloaded) and overused (overloaded), may emerge, leading respectively to either wastage of cloud resources or a decrease in services. So, the idea to incorporate meta-heuristic algorithms into TS has arisen so that complicated and varied incoming tasks (cloudlets) can be easily distributed within a reasonable time to available resources. The meta-heuristic techniques have shown themselves to be very able to solve scheduling problems, which are met by giving a detailed overview of traditional and heuristic approaches before deeply expanding into common cloud task meta-heuristic methods, then comprehensive systematic review, which includes new taxonomy and advantage of those methods. More specifically, major contribution keys of this study may be organized as follows: (1) (2) (3) (4)

Comprehensive study about cloud computing evolution and architecture Studying the mechanisms of existing load balancing. Providing an overview about task scheduling in cloud computing Identify important areas where new research can be performed with the optimization principle to better load balancing algorithms. (5) A systematic analysis is presented on cloud meta-heuristic TS. (6) Identifying research gaps and issues exist that can be further hurdle to researchers for LB algorithms The below is paper systematized: Section 2 elaborates the cloud computing concept with their architecture. It also consists of data center networking for VMs. At last, it talks about motivation of study. Section 3 elaborates the review of load balancing techniques, and task scheduling in CC in Sect. 4. Metaheuristic optimization algorithm comparison shows in Sect. 5. Section 6 explains related work also identifies some research gaps and issues and Sect. 7 concludes this work.

2 Cloud Computing: Architecture Web-based tools and technologies can be found in cloud computing resources. This allows the users to work remotely because the cloud can be used as “Internet”. Therefore, it is not processed as traditional outsourcing. It is also called Massive Computing. In this, the allocation of applications must be dynamic. No hardware or software has to be installed. Cloud computing aims to enable people who have no deep knowledge of information about all technology and applications [18].

A Comprehensive Study on Cloud Computing: Architecture, Load …

141

2.1 Overview “Cloud” is a virtualized reusable resource computing pool. It can change or handle a variety of workloads. The “Cloud’ term is a tarn of enormous infrastructure possessions, offering various utilities and hardware implementations as well as device software to end-users, alternatively enabling end-users to access their computing needs inconsistent way. The user needs no information about where to discover the device requirement and how the cloud operates. Cloud providers manage all obtainable tools to consumers where ‘computing’ is defined on an ‘SLA’. Cloud computing offers inevitable benefit in sharing over the internet of many configurable system properties and higher-level administrations that may be built with very little board effort. Cloud computing has been developed to promote the use of virtualization technology to allow end-users to use virtual services without infrastructure at a reasonable cost [19]. Cloud Computing Evolution in IT in shown in the Fig. 2. Now we are describing different modes of cloud computing as follows [21]: (1) Public Cloud: They are managed by CSP which owns facilities and data centers. Infrastructure is located on-site and companies may use pay-as-you-go and on-demand services. (2) Private Cloud: It is only created and operated by certain enterprises however 3rd party companies have control on behalf of the cloud owner to handle it. (3) Hybrid Cloud: It combines an only selection of all kinds of cloud deployments, such as public, private, or community cloud. Core operations are conducted in a private cloud, while a public cloud offers fewer basic resources. (4) Community Cloud: Several organizations or institutions of a shared purpose are sharing the community cloud. Universities that do this for learning and research are typical examples.

Fig. 2 Cloud computing evolution in IT [20]

142

S. Tiwari and C. Bhatt

2.2 Features of Cloud Computing Compared with other computing paradigms, CC provides a range of new features [22] and advantages. This section is briefly identified. • Scalability and On-Demand Services: Cloud computing connects customers on request with resources and services. Resources may be scaled across several data centers. • QoS (Quality of Service): In terms of CPU performance or hardware, memory capacity, and bandwidth, CC may provide QoS for users. • User-Centric Interface: Cloud interfaces are locational isolated and well-defined interfaces including web services and web browsers enable them to be accessed. • Autonomous System: CC applications are user-friendly, transparently controlled decentralized systems. But, software and data inside clouds may be reconfigured and merged automatically into a basic, user-specific platform. • Pricing: Cloud computing requires no investment at the start of the project. There is no need for capital expenses. Users can pay for services and capacities or pay for them, as they require them.

2.3 Architecture The applications, data, and utilities are all maintained in the cloud over the Internet and run applications, as well as stored data through the provision of software resources on-demand services in CC architecture. The back end and front end of a cloud architecture can be separated. The front end is available to the user by internet connections, enabling user interactions with the system [23]. The back end includes different cloud service models [24], like SaaS, PaaS, and IaaS. Often referred to as ‘Layered computing model’ is cloud computing architecture [25]. CC architecture can be classified into 4 layers that are hardware layer, infrastructure layer, platform layer, application layer as seen in Fig. 3. Description of each layer is defined as follows [26]: • Hardware Layer: Cloud is handled with physical resources. Controlling physical servers, switches, routers, the power system is the responsibility of the hardware layer. The implementation of the hardware layer is provided in a data center. There are several servers interconnected by routers and switches in the data center. Hardware layers have several issues including fault tolerance, hardware configuration, traffic management, and management of resources. • Infrastructure Layer: The virtualization layer is named as well. Cloud computing is an important aspect. Infrastructure layers focused on core aspects such as the use of virtualized technologies for the dynamic assignment of resources. The infrastructure layer uses virtualization techniques to collect processing and storage resources and divide physical resources. E.g. Xen, VMware.

A Comprehensive Study on Cloud Computing: Architecture, Load …

143

Fig. 3 Architecture of cloud computing

• Platform Layer: This layer is made up of operating system and application framework. It is installed on top of the infrastructure layer. The key principle of the platform layer is CC to reduce overhead for direct deployment in VM containers. E.g., Google App Engine runs on a platform layer to allocate API supports for data storage of various web applications. • Application Layer: It is built on the top level of cloud architecture. It is composed of an actual cloud application. Cloud applications have essential features to achieve better performance, lower operating cost, availability, and scalability. Thus, this architecture is more modular than other architecture (traditional architecture). Loosely coupled concepts are used in each layer. This architecture allows cloud computing to meet a vast variety of applications and to decrease total costs. No need for the high-power computer to run web-based applications is available in the cloud computing infrastructure.

2.4 Data Center in Cloud Network Cloud migration platforms are a new and rapid trend. Cloud offers a standardized front-end interface, enabling a large number of applications to be executed on a

144

S. Tiwari and C. Bhatt

similar hardware platform. DCs are the backbone of cloud computing development networks. Cloud output depends on the data centers’ computing, storage, and network availability [27]. The demands for cloud working load in today’s data centers, thus, is reflected in demands for resources. The data center consists of servers, devices and storage, cooling systems, network devices, power systems, etc. [28]. DCs are for large-scale systems for services like online enterprises, smart grid and mathematical computations. DC is being used to model core services on the cloud infrastructure network. It comprises a group of hosts who handle a group of VM whose roles are to handle “low-level” processing, and a minimum of 1 DC should be set up to commence simulation. Virtualization technologies [29, 30] are additionally supplied to data centers which allow multiplexed and shared multiple resources between huge no. of users by various time-varying access patterns [31]. DC management is difficult. There are approximately two types of challenge: (1) Management of resources, which focuses on dynamic workloads management provided a resource pool and (2) Planning of capacity, which concentrates on the provision of resources. DCN (Data Center network) comprises DC also offers data center connections, as defined in its networking topology, the routing and switching equipment as well as the protocols it uses [32]. for the following purposes DCN provides some features to better organize cloud computing [33]: • DCN allows thousands of data center servers to be efficiently connected such that cloud computing can extend its operation easily by adopting the DCN topology. • In massive machine-to-machine connectivity, DCN provides traffic reliabilities and efficiencies that generate activities of cloud computing as working loads distributed on servers at data centers. • DCN embraces different techniques of virtualization that help DCN build virtual machines (VMs), virtual networks, and virtual functions. The DCN should be scalable to isolate and migrate to large numbers of virtual instances. • Existing DCN research has produced applications in some use cases, including green computing and DC backup, that may also address challenges of cloud infrastructure.

2.5 Virtual Machines in Cloud Computing Cloud computing includes parallel processing principles and distributed computing to provide shared services via physical server hosting VMs. The service-orientated architecture reduces connectivity costs to collect customer information deliver more flexibility and demand-driven services, etc. The premise behind the generated cloud computing concept is that the processing of information is a public service, which can be achieved more effectively in massive computer farms and storage systems that can be made available worldwide through the Internet. Effective management of

A Comprehensive Study on Cloud Computing: Architecture, Load …

145

Fig. 4 Process of VM scheduling [35]

VMs [34] specifically affects the use of the system’s resources and QoS. The amount of VM distributing across physical servers seems to be inconsistent over a certain period, given the dynamic nature of the cloud environment. In this case, VMs have to be moved from an overwhelmed server to a load-based server to balance the load. Figure 4 presents the general cloud data center VM scheduling process. Scheduling can be widely categorized into user applications, control of resources, and scheduling mode. VM’s component handles the allocation of various hosts to various virtual machines so that the computing cores can be allocated (by the host) to VMs. This configuration is based on method, and the default VM assignment policy is ‘firstcome, first-serve.’ Virtualization allows the live migration of VMs [36] with several VMs loaded on several physical machines (PMs) referred to as VMs. An efficient strategy to reduce energy consumption, operational costs, hardware costs, conformance/violation of SLAs, CO2 emissions also enhance hardware and service reliability, efficiency and hardware life, LB and use of a CCS may be a VM consolidation algorithm. VM consolidation in the cloud environment can essentially reduce energy use and QoS.

146

S. Tiwari and C. Bhatt

2.6 Motivation of Study The modern cloud computing model offers sophisticated benefits and benefits compared to previous computing paradigms and is adapted, migrated, and adopted by many organizations. Cloud computing has evolved from a prospective logic over the past few years; business is the idea of virtualization to a rapidly increasing sector in the IT sector. Today, recession-hit firms are becoming increasingly aware that they can easily tap into the cloud and quickly access best-of-breed business applications or drastically enhance their infrastructure services at marginal costs. But, some issues, difficulties and effects are still found which scholars, academics and business intelligence (BI) practitioners are presently addressing. Clouds still present enterprises with security concerns [37] by using cloud computing. Upon details and IT critics behind the firewall, users are still concerned about the susceptibility to attack. Computing in clouds also provides reliability around the clock. There were some cases of outages for a few hours from cloud computing services. Contrary to the conventional computing model, cloud computing uses virtual computing technologies, as users have access to cloud computing services, they can leak hidden information. Attackers will evaluate crucial jobs based on the user’s computing task [38]. The growth of cloud computing needs to open standards. The interpretation of most cloud providers with an [39] API, usually welldocumented but often special to its execution and thus not interoperable. Heavy transaction-oriented also other applications of data-intensive in which CC can miss sufficient efficiency can be the key to performance. In addition, users far from cloudbased services suffer from high latency and delay. Software and hardware can be saved for companies offering the CC; however, higher latency costs can be incurred. Bandwidth costs may be small for smaller internet applications and not resourceintensive, nonetheless data-intensive applications can rise significantly. Users can be certain that even certain cloud computing services providers can never become irrelevant, or get bought and swallowed up by some bigger company. “Cloud potential providers can retrieve the data and it is imported into replacement application through either format” -Gartner [40]. Cloud service providers are required to comply with legal issues concerning data responsibility and ownership for loss or misuse of data. Legal problems vary from those caused by traditional hosting or outsourcing [41]. These problems and challenges encourage us to focus more on this subject of cloud computing problems.

3 Load Balancing in Cloud Computing LB is a key problem and issue in cloud environments [42]. The method of allocating and reassigning the load between usable resources is designed to optimize performance, minimize cost and response times, improve efficiency, resource use and save energy. Excellent load balancing strategies can include SLA and user satisfaction.

A Comprehensive Study on Cloud Computing: Architecture, Load …

147

Therefore, it is a prerequisite to the performance of CC environments to ensure reliable algorithms and processes for load balancing. LB model is demonstrated in Fig. 5, LB accepts user requests and executes load balancing algorithms to allocate applications to VMs. Load balancer determines that VM’s next request must be allocated. The controller of DC is responsible for task management. LB algorithm provides tasks to delegate tasks to appropriate VM. VM manager is responsible for VMs. Summary of LB policies are shown in Table 1.

Fig. 5 Load balancing model [43]

Table 1 Summary of LB policies S. No. Location policy

Information policy

Selection policy

Transfer policy

1

Find the right partner for transfer

Determine how long Selecting factors for Includes: task information about transferring a task: re-scheduling task nodes must be migration overhead migration collected

2

Task monitor availability of resources needed for partner migration

There are three A no. of remote kinds: demandsystem calls time of driven, periodic, and task execution state policies information policy

Depends upon thresholds in terms of load units

148

S. Tiwari and C. Bhatt

Virtualization is leading CC technology. Virtualization aims mostly to share costly hardware between VMs. VM is computer software execution on which operating systems and applications will work. The user requests are processed by VMs. Users are found around the globe and send their requests arbitrarily. For processing, requests must be delegated to VMs. The distribution of tasks is also an important problem in CC. If several VMs are overloaded when others are idle or have some work to do, QoS is reduced. By reducing QoS, consumers are not happy with the system and will never return. The VM is created and managed by a hypervisor or VMM (Virtual Machine Monitor). Multiplexing, suspension (storage), provision (resume) and life migration [44] are four operations of VMM. For load balancing, certain operations are required. In [45] it was specified that load balance would take account of two tasks: allocation of resources and scheduling of tasks. The effect of these 2 tasks is high infrastructure availability, increased use of power, energy savings, reduced cost of resource use, maintaining cloud storage elasticity, and reducing carbon emissions.

3.1 Classification of LB Techniques This section describes basic testing activities using various approaches to load balancing. The study of the various methods of load balancing for edge computing is explained as follows [46]. Different forms of strategies of LB are seen in Fig. 6. Different types are introduced to provide an improved balance of loads in edge computing, like security, traffic-load based, heterogeneous, optimization-based, joint-load based, heuristic and multi-access based, dynamic load-based, allocation-based, and allocation-based approaches.

3.2 Load Balancing Metrics The metrics for LB in CC are reviewed in this section and summarized as follows [47]: • Response time: It calculates the overall time required to fulfill a submitted task by the system. • Throughput: This measurement is utilized to quantify the number of processes per unit time performed. • Scalability: An algorithm is capable of uniform LB in the system as per demands, as no. of nodes increases. It is a highly scalable preference algorithm. • Makespan: This metric is utilized to measure maximum completion time or the amount of time a user has attributed the tools. • Migration time: time to switch tasks from overloaded node to undercharged node.

A Comprehensive Study on Cloud Computing: Architecture, Load …

149

Fig. 6 Categorization of LB techniques in CC

• Fault tolerance: It decides the algorithm’s capacity to perform LB in event of such nodes or ties breakdown. • Performance: After the LB algorithm, it tests system efficiency. • Degree of imbalance: This measurement the imbalance of VMs. • Carbon emission: The amount of carbon that all materials generated is calculated. Load balancing played an important part in reducing this metric by switching and shutting loads from underloaded nodes. • Energy consumption: It determines how much energy all nodes consume. Load balancing prevents overheating and thereby reduces energy consumption by LB across all nodes.

4 Task Scheduling in Cloud Environment Cloud computing handles a range of virtualized capabilities that make planning an important component. A customer can use thousands of virtualized properties in the cloud for each task [48]. Manual scheduling is therefore not a workable alternative. The underlying concept behind scheduling consists of distributing tasks (complex and diverse nature) between cloud resources in a manner that reduces time losses and maximizes efficiency by scheduling algorithms. Several study activities have in the past looked at task planning. Cloud operating resources are tracked and loads are computed for any resource before assigning workload or tasks on VMs (resources). If any VM is in over-used mode, a task for those types of resources is not assigned. The task is a workpiece to be carried out within a given time frame. Inside the cloud, the task is in two forms [49]: a task that is independent and dependent. Task

150

S. Tiwari and C. Bhatt

scheduling is the mechanism by which the services are assigned to this task for a certain amount of time. TS is a very common theme in the field of CC. It is used to plan tasks to improve the usage of resources by the allocation of such tasks to definite resources in particular. The main objective of the task algorithm is to increase reliability, enhance service quality, maintain productivity across tasks, and reduce costs. In the TS process, virtual tools are used to their maximum potential. For high performance, effective resource scheduling is essential. The completion period and the cost of completing the task are two different criteria.

4.1 Types of Task Scheduling Categories of TS are as follows [50]: • • • • • •

User-Level Scheduling Cloud Service Scheduling Heuristic Scheduling Static and Dynamic Scheduling Workflow Scheduling Real-Time Scheduling.

4.2 Framework of TS in CC Environment There is a framework [51] for the TS system in CC. In Fig. 7, users are accessing the cloud environment through the internet. The cloud part shows how the cloud is managing to serve various requests given by the consumers. (1) SLA Monitor: First, a client sends the service request. SLA monitor then reviews the requested QoS application before deciding whether the submission should be accepted or denied. SLA Monitor is responsible for monitoring the success of the job submitted and disciplinary measures must be taken promptly if there are any violations. (2) Resource Discovery and Monitoring: Resource search can essentially be defined as the job of the manufacturer to locate the right resources to satisfy incoming customer demands. Cloud computing’s main advantages are the ability to receive and distribute data on demand, so the management of resources must be continuous. (3) Task Scheduling: The input of the algorithm for task scheduling is usually an abstract model. This abstract model describes tasks without defining were tools to perform the tasks are physically placed. (4) Reschedule: If a task cannot be done due to a processor malfunction or other problem, the incomplete task in the next calculation will be rescheduled.

A Comprehensive Study on Cloud Computing: Architecture, Load …

151

Fig. 7 Framework of TS in CC environment

(5) Scheduling Optimizer: Several relevant candidates are illuminated after the information is obtained about the available resources in the cloud. The framework for resource selection chooses a solution that meets all demands and optimizes infrastructure usage. The allocation of resources can be made using an algorithm for optimization. Various optimization methods can be used by common and well-known techniques, including simple metaheuristic algorithms. Examples include GA, ACO and PSO for the cloud. (6) Advanced Resource Reservation Monitor: The Advanced Resource Reservation Control ensures QoS connections in the data center to different services. In the future, users will protect the necessary services. This is necessary if some processors need to be run to complete critical time applications such as realtime and Workflow applications for parallel applications. Providers can more reliably forecast potential demand and use. (7) Data Centers and VMs: Data centers are the infrastructure or hardware part where all the physical servers present. Physical servers are not used as viable

152

S. Tiwari and C. Bhatt

users. Through virtualization, they are converted into different VMs and users’ jobs are going to run on these VMs in scheduling tasks.VM manager manages the VMs.

5 Meta-Heuristic Optimization Techniques Due to the usefulness of these metaheuristic algorithms in solving massive computational and complex problems [52] has been enormously common in the past 20 years. There are some useful metaheuristic characteristics such as: (1) These algorithms are not unique to problems. (2) Effectively explore a meta-heuristic algorithm in search areas to find (near) optimal solution or sub-optimal solution of NP-complete problems. (3) Approximate and typically non-deterministic meta-heuristic algorithms. Metaheuristic algorithms are problem-independent and can be used to solve issues in a variety of domains. Metaheuristic approaches are standard techniques for solving NP-hard optimization problems. Meta-heuristic = Heuristic + Randomization. Many meta-heuristic algorithms are available in the cloud environment to find an approximate (suboptimal) solution to an NP-Complete problem in a short amount of time. TS is one of the problems due to the vast space for a solution which takes a long time to find an optimal solution. Diverse choice of meta-heuristic algorithms like PSO, ACO, GA, Artificial Bee Colony (ABC), Simulated Annealing (SA), Differential evolution algorithm (DEA), BAT optimization, Bacteria Foraging Optimization, Firefly optimization, Cat swarm optimization, Lion optimization, Cuckoo search, etc. We also addressed and analyzed numerous task scheduling algorithms in the cloud environment with a meta-heuristic method and their benefits and drawbacks (demonstrated in Table 2). In evaluating precision solutions to a specific problem, the right choice of the optimization algorithm is greatly helpful. At first, we will analyze the PSO algorithm since it is easier to converge the PSO algorithm and the complexity than other metaheuristic algorithms.

6 Literature Survey 1.

Oliveira et al. [3] Recently, a great many projects have carried out extensive cloud research, including load balancing and task scheduling. These works illustrate some of the core problems facing state-of-the-art load balancing

Author (Year)

Elaziz et al.

Adhikari et al.

Sharma and Kumar PSO-COGEN

Alaaeldin and Almezeini

S. No.

1

2

3

4

Lion algorithm

PSO-COGENT algorithm

BAT

Hybrid algorithm

Algorithm

Independent

Independent

Independent

Independent

Nature of tasks

Enhance cost, makespan time, resource utilization, reliability, etc

Enhance makespan system time and throughput

Benefits

TS by lion

Enhance cost, makespan time, resource usage and imbalance degree

Modified PSO Enhance time, depends on APSO-VI throughput, energy consumption, QoS, cost parameters, etc

LB-RC algo via BAT

MSDE algorithm (MSA with DE)

Techniques

Table 2 Summary of the tested meta-heuristic algorithms, including their benefits and drawbacks

(continued)

Algorithm considers no limit and did not address the compromise solution

QoS parameters such as availability, reliability, etc. cannot be taken into consideration. VMs can be overloaded when applications of VMs have been improperly described

Algorithm doesn’t discuss compromise among time and cost QoS parameter

A single objective (focused only time) is the algorithm performed and does not take account of other QoS parameters such as reliability, cost, energy

Drawbacks

A Comprehensive Study on Cloud Computing: Architecture, Load … 153

Author (Year)

Addya et al.

Zamanifar and Rastkhadiv

Lin et al.

Pacini et al.

S. No.

5

6

7

8

Table 2 (continued)

ACO

Binary PSO

ABC

SA

Algorithm

Independent

NA

Independent

Independent

Nature of tasks

Benefits

2 Level cloud scheduler method by ACO algorithm

Modified sigmoid transfer function

Agile task handling method via ABC

Enhance response time, throughput and makespan time

Enhance time of High utility itemset mining

Enhance makespan time and degree of imbalance

MVMP algorithm via enhances, energy SA consumed, profit and execution time

Techniques

(continued)

Algorithms do not take into account the time and tasks priority; indirect communication mechanisms are utilized to communicate information (pheromone)

The sigmoid function proposed would not account for the last step of the operation

Algorithms are only used to carry out independent tasks and main performance parameters such as reliability, energy, and cost, etc. are not considered, where the deadline is a restraint

The solution to static VM migration takes time

Drawbacks

154 S. Tiwari and C. Bhatt

Author (Year)

Kaushal and Verma

Ramezani and Khadeerhussain

Jana and Kulia

Krishna and Babu

Huo and Zhan

S. No.

9

10

11

12

13

Table 2 (continued)

Algorithm

Hybrid PSO

ABC

TBSLB- PSO

PSO based scheduling algorithm

PSO based scheduling

Nature of tasks

Independent

Independent

Independent

Independent

Workflow applications

Techniques Reduce execution costs and time at the same time

Benefits

Energy usage and other effective QoS are not taken into account

Drawbacks

Time of response, runtime and performance, number of migrated tasks, throughput

Reduce task transfer time and makespan time

SA with PSO is added Enhance algorithm search capabilities with SA

Task migration based LB using ABC

Task migration from overloaded VMs to under-loaded VMs

After adding SA and PSO the complexity of the algorithm is improved

Only non-preemptive tasks and deadline of tasks are not considered for proposed algorithm

Performance of algorithm suggested is not related by other state-of-the-art algorithms

task migration, Cloud Enhance rejection ratio Gbest is affected by resource broker time, expense and task local minima and method energy utilization is not taken into account

Bi-objective priority-based PSO

A Comprehensive Study on Cloud Computing: Architecture, Load … 155

156

2.

3.

4.

5.

6.

7.

8.

S. Tiwari and C. Bhatt

applications. Several task scheduling mechanisms have been defined by many researchers for load balancing. Jena [53], focused on TSPSO (task scheduling using multi-objective nested particle swarm optimization) to enhance energy and processing time. Outcome from TSPSO was simulated through cloud platform for open source applications (CloudSim). The findings were finally related to the existing scheduling algorithms and found an optimum balance outcome for different goals with the proposed algorithm (TSPSO) [54]. Wang and Zhou [55], Proposed a technique for solving the MapReduce load imbalance issue created by the use of the default partition algorithm of the Hadoop platform. This suggested approach will optimize the activities and balance inputs of a decrease process in the map phase by using multiple partition techniques. Moreover, this suggested technique will use idle nodes to completely offset high load nodes, to attain optimum work scheduling throughout the reduction phase execution method [56]. Velde and Rama [57], In this cloud partitioning strategy, LB and resource optimization are effectively extended. Besides the optimal time of refresh that sets the state adjustments of the data center partitions for efficient use of resources. CloudSim is a simulation platform utilized to create prototype applications to prove hypothesis proof. Results have shown that this approach increases cloud resource efficiency, performance, and optimization [53]. Basha and Padmavathi [58], proposed an innovative, dynamic, and elasticity algorithm perform LB by ACO to perform load Balance between Systems existing in DCs [55]. Vijayakumar and Kanthimathi [59], The suggested scheme used additional virtual machines using genetic methods to address the requests of the highest virtual machines. The allocation of the best virtual machines will deal with demands very efficiently and quickly. During execution, the load might be balanced via the ACO technique if those VMs were overloaded by requirement. The above strategy will share the extra charge with other virtual machines, gently loaded and idle. On the other hand, after their work completion or in idleness total energy consumption is enhanced by turning away VMs [57]. (MPSO) hybridization and an improved Q-learning algorithm called QMPSO. Hybridization is done to change MPSO velocity by pbest and gbest depends upon the best action provided by improved Q-learning. The objective of hybridization is to improve machine performance by balancing the charge among VMs, optimize VM’s output and maintain the balance between task priorities by maximizing the time of work. By comparing QMPSO results obtained during the simulation process with the existing LB and scheduling algorithm, the robustness of this algorithm is evaluated. The simulation compared with real results of the platform reveals that the proposed algorithm outperforms its competitor [60]. Malik and Bansal [61], An optimized approach to scheduling model and resource cost scheduling model labeled MultiFaceted Optimization Scheduling Framework (MFOSF) was timely in this study to resolve the restrictions and

A Comprehensive Study on Cloud Computing: Architecture, Load …

157

change solution quality. The resource Cost model shows the relation between customer budget and production costs during scheduling. PSO can be used to achieve a model focused on optimizing efficiency and cost. There have been several simulations to test this approach using 4 different metrics (a) Makespan (b) Cost (c) Resource Utilization (d) Deadline. Based on the above parameters, experimental results demonstrate that the MFOSF-PSO approach was better than other models and in the best-case scenario increased by 57.4% [61]. 9. Devaraj et al. [62], As a firefly hybrid and Improved Multi-objective Particle Swarm Optimization (IMPSO) technology, abbreviated as FIMPSO, the latest LB algorithm is proposed for use. This technique uses FF (Firefly) algorithm to limit search space to detect the improved response using the IMPSO technique. The proposed FIMPSO algorithm generated and improved effective average load for crucial measures, including proper use of resources and response time for tasks. The simulation result has shown the efficient performance of the FIMPSO model suggested compared to other methods. From the resulting simulation, the FIMPSO algorithm is understood to have generated a successful outcome with an average response time of 13.58 ms maximum, overall CPU usage of 98%, 93% memory utilization, 67% reliability and 72% throughput, along with 148, which were superior to all other techniques compared [62]. 10. Miao et al. [63], The new algorithm called adaptive Pbest Discrete PSO (APDPSO) for PSO-based static load balancing was proposed to combat this problem. Good solution held in the external archive is used to upgrade the particles’ personal best positions and to update velocity and direction vectors of particles by proposing a probability and comparability approach for PSO. MATLAB and CloudSim systems perform simulation experiments with random synthetic tasks. The findings revealed that our proposed algorithm dramatically increased swarm convergence and diversity and reduced the level of load imbalance relative to state art in the field [63].

6.1 Research Gaps and Issues This section addresses the study deficiencies of cloud computing that need to be addressed in terms of existing load-balancing techniques. • Lightweight security solutions were not considered for increased LB efficiency [64]. • Method [65] failed to implement the paradigm built-in CC and real field of joint edge computing. The procedure failed to increase QoE and reduce incoming traffic based on the adaptive bitrate approach for mobile video streaming and optimized edge caching [66]. • The cross-layer approach on heterogeneous networks was more complicated [67]. • The approach in [68] neglected to take into account MEC server security levels, which is very important to industry data.

158

S. Tiwari and C. Bhatt

• In [69], the complexity was not properly calculated and the reliability was not improved. In [70] the strategy to balance the positive and negative impacts on the migration of services has not been established. • The method in [71] needed additional major structures for orchestration to improve the system performance. • Performance has been improved, but the allocation of vehicles was highly imbalanced, leading to inefficient QoS for a vehicle to infrastructure (I2V) communication [72]. • However, the forecast was inaccurate because the task length was smaller and the training set was not included [73]. Large-scale heterogeneous MEC problems for optimizing training speed were not taken into consideration [74]. • In [75], live video and car navigation services could not be used. No further data sets were considered to improve the system’s performance [76]. • The load-aware user associations that contain dynamic settings were not taken into account [77]. The transmitting power and rates of SCBSs were not taken into account in [78] to achieve minimum energy consumption. • Because of many interferences and noises, the communication channel was not perfect [79]. They are normally unable to understand the degradation of QoE [80]. The process of the reorder did not however alter the efficiency of the load balancing [81]. The challenges of these approaches are taken into account as a reason for developing a new load-balancing method.

7 Conclusion AC is the product of the Virtualization concept-enabled grid computing. LB is a challenge in the cloud environment, as it is a complex task to share the task equally between VMs. LB is the main CC mechanism that prevents overloading nodes. Load balancing stabilizes service quality (QoS), covering response time, cost, throughput, efficiency, and the use of resources. At the height of time, servers cannot manage incoming requests with no. of VMs available, meaning that some additional virtual machines needed to be run without fault or delay. The benefit of switching to a virtual environment is important, but there are significant efficiency and cost optimization issues in cloud computing due to different types of need for resources and multiple tasks. The success standard and requirements of SLA to be retained therefore become very difficult because of these constraints. Various alternatives to the planning model and resource cost timeline model have been applied to solve these limitations and to change solution quality. Different meta-heuristic optimization strategies were applied to conduct TS in the CC environment to manage demand on virtual machines. This survey will offer an idea of the newest developments in LB in cloud systems and will provide an analysis of the technological advances, company benefits and growing use of data centers, and future trends. Amazon EC2, Microsoft Azure, and

A Comprehensive Study on Cloud Computing: Architecture, Load …

159

Google App Engine are compared. This article highlights research gaps and load handling problems in CC andsummarizes the superiority of CC. The future analysis is to find effective methods for calculating the cloud results.

References 1. Academic paper http://www.mcs.csueastbay.edu/~lertaul/Cloud%20Security%20CamR EADY.pdf 2. Chen G, Lu J, Huang J, Wu Z (2010) SaaAS—the mobile agent-based service for cloud computing in internet environment. In: Sixth international conference on natural computation, ICNC 2010. IEEE, Yantai, Shandong, China, 2010, pp 2935–2939. ISBN: 978-1-4244-5958-2 3. Oliveira D, Baião F, Mattoso M (2010) Towards taxonomy for cloud computing from an e-science perspective. In: Cloud computing: principles, systems, and applications (to be published). Springer, Heidelberg 4. Singh G, Sood S, Sharma A (2011) CM-Measurement facets for cloud performance. IJCA 23(3). Lecturer, Computer Science and Engineering, Eternal University, Baru Sahib (India) 5. Ertaul L, Singhal S (2009) Security challenges in cloud computing. California State University, East Bay 6. Global Netoptex Incorporated.—Demystifying the cloud. Important opportunities, crucial choices, pp 4–14. Available: http://www.gni.com. 13 Dec 2009 7. Gulshan S, Kalra M (2014) A novel approach for load balancing in cloud data center. 978-14799-2572-8/14/$31.00 c_2014. IEEE 8. Wu H, Ding Y, Winer C, Yao L (2010) network security for virtual machines in cloud computing. In: 5th International conference on computer sciences and convergence information technology, pp 18–21, Seoul, Nov 30–Dec 2, 2010. ISBN: 978-1-4244-8567-3 9. Kliazovich D, Arzo ST, Granelli F, Bouvry P, Khan SU (2013) eSTAB: energy-efficient scheduling for cloud computing applications with traffic load balancing. In: Green computing and communications (GreenCom). IEEE, pp 7–13 10. Greenberg A, Hamilton J, Maltz DA, Patel P (2008) The cost of a cloud: research problems in data center networks. ACM SIGCOMM Comput Commun Rev 39(1):68–73 11. Lu X, Kong F, Yin J, Liu X, Yu H, Fan G (2015) Geographical job scheduling in data centers with heterogeneous demands and servers. In: 2015 IEEE 8th international conference on cloud computing, pp 413–420. https://doi.org/10.1109/CLOUD.2015.62 12. Armbrust M, Fox A, Griffith R, Joseph AD, Katz R, Konwinski A, Lee G, Patterson D, Rabkin A, Stoica I et al (2010) A view of cloud computing. Commun ACM 53(4):50–58 13. Wang W, Li B, Liang B (2013) Dominant resource fairness in cloud computing systems with heterogeneous servers. arXiv preprint arXiv:1308.0083 14. Shafiq DA, Jhanjhi NZ, Azween Abdullah (2021) Load balancing techniques in cloud computing environment: a review. J King Saud Univ Comput Inf Sci. https://doi.org/10.1016/ j.jksuci.2021.02.007 15. Khan S, Sharma N (2014) Effective scheduling algorithm for load balancing (SALB) using Ant Colony 16. Optimization in cloud computing. Int J Adv Res Comput Sci Softw Eng 4(2) 17. Kushwah P (2014) A survey on load balancing techniques using ACO algorithm. (IJCSIT) Int J Comput Sci Inf Technol 5(5):6310–6314 18. Farrag AAS, Mahmoud SA, EI Sayed M, EI- Horbaty (2015) Intelligent Cloud Algorithms for Load Balancing problems: a survey. In: 2015 IEEE seventh international conference on intelligent computing and information systems (ICICIS ‘2015) 19. Houssein EH, Gad AG, Wazery YM, Suganthan PM (2021) Task scheduling in cloud computing based on meta-heuristics: review, taxonomy, open challenges, and future trends. Swarm Evol Comput 62:100841. https://doi.org/10.1016/j.swevo.2021.100841

160

S. Tiwari and C. Bhatt

20. Chen Y, Paxson V, Katz RH (2010) What’s new about cloud computing security. In: University of California, Berkeley Report No. UCB/EECS-2010-5, January 2010 21. George SS, Suji Pramila R (2021) A review of different techniques in cloud computing. Mater Today Proc. https://doi.org/10.1016/j.matpr.2021.02.748 22. Radarnetworks and Novaspivak. http://radarnetworks.com 23. Odun-Ayo I, Ananya M, Agono F, Goddy-Worlu R (2018) Cloud computing architecture: a critical analysis. In: 2018 18th international conference on computational science and applications (ICCSA), 2018, pp 1–7. https://doi.org/10.1109/ICCSA.2018.8439638 24. Prasad MR, Lakshman Naik R, Bapuji V (2013) Cloud computing: research issues and implications. Int J Cloud Comput Serv Sci (IJ-CLOSER) 2(2):134–140 25. Verma, Kaushal S (2011) Cloud computing security issues and challenges: a survey. (July):445– 454 26. Tiwari K, Chaudhary S, Shanu K (2015) Survey paper on cloud Computing. In: International conference on emerging trends in technology, science and upcoming research in computer science, Apr 2015, pp 1777–1782 27. Zhang Q, Cheng L, Boutaba R (2010) Cloud computing: state-of-the-art and research challenges. J Internet Serv Appl I:7–18 28. Joshi S, Kumari U (2016) Cloud computing: architecture and challenges. Mody Univ Int J Comput Eng Res 1(1):56–60 29. Birke R, Chen LY, Smirni E (2012) Data centers in the cloud: a large-scale performance study. In: 2012 IEEE Fifth international conference on cloud computing, pp 336–343. https://doi.org/ 10.1109/CLOUD.2012.87 30. Abts D, Felderman B (2012) A guided tour through data-center networking. Queue 10(5):10:10–10:23 31. Wood T, Tarasuk-Levin G, Shenoy PJ, Desnoyers P, Cecchet E, Corner MD (2009) Memory buddies: exploiting page sharing for smart colocation in virtualized data centers. In: VEE, 2009, pp 31–40 32. Mehta S, Neogi A (2008) ReCon: a tool to recommend dynamic server consolidation in multicluster data centers, In: NOMS, 2008, pp 363–370 33. Chen Y, Das A, Qin W, Sivasubramaniam A, Wang Q, Gautam N (2005) Managing server energy and operational costs in hosting centers. In: SIGMETRICS, 2005, pp 303–314 34. Wanga B, Qi Z, Maa R, Guana H, Vasilakos AV (2015) A survey on data center networking for cloud computing. Comput Netw 91:528–547 35. Wang A, Iyer M, Dutta R, Rouskas GN, Baldine I (2013) Network virtualization: technologies, perspectives, and frontiers. J Lightwave Technol 31(4):523–537 36. Sahu Y, Agrawal N (2015) A survey paper: cloud computing and virtual machine migration. IJCSN J 4(4):577–581 37. Liu L, Qiu Z (2016) A survey on virtual machine scheduling in cloud computing. In: 2016 2nd IEEE international conference on computer and communications (ICCC), 2016, pp 2717–2721. https://doi.org/10.1109/CompComm.2016.7925192 38. Zolfaghari R, Sahafi A, Rahmani AM, Rezaei R (2021) Application of virtual machine consolidation in cloud computing systems. Sustaine Comput Inf Syst 30:100524. https://doi.org/10. 1016/j.suscom.2021.100524 39. Mills E (2009) Cloud computing security forecast: clear skies 40. Jiang J, Wen W (2010) Information security issues in cloud computing environment. Netinfo Secur. https://doi.org/10.3969/j.issn.1671-1122.2010.02.026 41. Clark C, Fraser K, Hand S, Hansen JG, Jul E, Limpach C, Pratt I, Warfield A (2005) Live migration of virtual machines. In: Proceedings of NSDI’05. Berkeley CA, USA, 2005. USENIX Association, pp 273–286 42. Gartner, Seven cloud-computing security risks. http://www.infoworld.com. 02 July 2008 43. Prasad MR, Gyani J, Murti PRK (2012) Mobile cloud computing implications and challenges. IISTE J Inf Eng Appl (JIEA) 2(7):7–15. http://iiste.org 44. Jadeja Y, Modi K (2012) Cloud computing—concepts, architecture and challenges. In: International conference on computing, electronics and electrical technologies [ICCEET]

A Comprehensive Study on Cloud Computing: Architecture, Load …

161

45. Gupta H, Sahu K (2014) Honey bee behavior based load balancing of tasks in cloud computing. Int J Sci Res 3(6) 46. Hwang K, Dongarra J, Fox GC (2013) Distributed and cloud computing: from parallel processing to the Internet of Things 47. Ivanisenko IN, Radivilova TA (2015) Survey of major load- balancing algorithms in distributed system. In: Information technologies in innovation business conference (ITIB) 48. Pydi H, Iyer GN (2020) Analytical review and study on load balancing in edge computing platform. 2020 Fourth international conference on computing methodologies and communication (ICCMC), 2020, pp 180–187. https://doi.org/10.1109/ICCMC48092.2020.ICCMC-00036 49. Jafarnejad Ghomi E, Masoud Rahmani A, Nasih Qader N (2017) Load-balancing algorithms in cloud computing: a survey. J Netw Comput Appl 88:50–71. https://doi.org/10.1016/j.jnca. 2017.04.007 50. Arunarani AR, Manjula D, Sugumaran V (2018) Task scheduling techniques in cloud computing: a literature survey. Future Gener Comput Syst. https://doi.org/10.1016/j.future. 2018.09.014 51. Abdul Qadir OS, Ravi G (2020) A survey on task scheduling algorithms in cloud computing. Int J Innovations Eng Technol 15(4):29–35. https://doi.org/10.21172/ijiet.154.06 52. Sharma P, Shilakari S, Chourasia U, Dixit P, Pandey A (2020) A survey on various types of task scheduling algorithm in cloud computing environment. Int J Sci Technol Res 9(01):1513–1521 53. Jena RK (2015) Multi objective task scheduling in cloud environment using nested PSO framework. Procedia Comput Sci 57:1219–1227. https://doi.org/10.1016/j.procs.2015.07.419 54. Venu G, Vijayanand KS (2020) Task scheduling in cloud computing: a survey. Int J Res Appl Sci Eng Technol (IJRASET) 8(V):2258–2266 55. Wang S, Zhou H (2016) The research of MapReduce load balancing based on multiple partition algorithm. In: 2016 IEEE/ACM 9th international conference on utility and cloud computing (UCC), pp 339–342 56. Kumar M, Sharma SC, Goel A, Singh SP (2019) A comprehensive survey for scheduling techniques in cloud computing. J Netw Comput Appl. https://doi.org/10.1016/j.jnca.2019. 06.006 57. Velde V, Rama B (2017) Simulation of optimized load balancing and user job scheduling using CloudSim. In: 2017 2nd IEEE international conference on recent trends in electronics, information and communication technology (RTEICT), pp 1379–1384. https://doi.org/10.1109/RTE ICT.2017.8256824 58. Padmavathi M, Basha SM (2017) Dynamic and elasticity ACO load balancing algorithm for cloud computing. In: 2017 International conference on intelligent computing and control systems (ICICCS), pp 77–81. https://doi.org/10.1109/ICCONS.2017.8250571 59. Kanthimathi M, Vijayakumar D (2018) An enhanced approach of genetic and ant colony based load balancing in cloud environment. In: 2018 International conference on soft-computing and network security (ICSNS), pp 1–5. https://doi.org/10.1109/ICSNS.2018.8573608 60. Kumar KP (2018) gravitational emulation-grey wolf optimization technique for load balancing in cloud computing. In: 2018 Second international conference on green computing and Internet of Things (ICGCIoT), pp 177–184. https://doi.org/10.1109/ICGCIoT.2018.8753108 61. Bansal M, Malik SK (2020) A multi-faceted optimization scheduling framework based on the particle swarm optimization algorithm in cloud computing. Sustain Comput Inf Syst 28:100429. https://doi.org/10.1016/j.suscom.2020.100429 62. Devaraj AFS, Elhoseny M, Dhanasekaran S, Laxmi Lydia E, Shankar K (2020) Hybridization of firefly and improved multi-objective particle swarm optimization algorithm for energy efficient load balancing in cloud computing environments. J Parallel Distrib Comput 142:36–45. https:// doi.org/10.1016/j.jpdc.2020.03.022 63. Miao Z, Yong P, Mei Y, Quanjun Y, Xu X (2021) A discrete PSO-based static load balancing algorithm for distributed simulations in a cloud environment. Future Gener Comput Syst 115:497–516. https://doi.org/10.1016/j.future.2020.09.016 64. Li J, Luo G, Cheng N, Yuan Q, Wu Z, Gao S, Liu Z (2018) An end-to-end load balancer based on deep learning for vehicular network traffic control. IEEE Internet of Things J 6(1):953–966

162

S. Tiwari and C. Bhatt

65. Dong Y, Xu G, Ding Y, Meng X, Zhao J (2019) A ‘joint-me’task deployment strategy for load balancing in edge computing. IEEE Access 7:99658–99669 66. Liu J, Shou G, Liu Y, Hu Y, Guo Z (2018) Performance evaluation of integrated multi-access edge computing and fiber-wireless access networks. IEEE Access 6:30269–30279 67. Liu L, Chan S, Han G, Guizani M, Bandai M (2018) Performance modeling of representative load sharing schemes for clustered servers in multiaccess edge computing. IEEE Internet of Things J 6(3):4880–4888 68. Niu X, Shao S, Xin C, Zhou J, Guo S, Chen X, Qi F (2019) Workload allocation mechanism for minimum service delay in edge computing-based power Internet of Things. IEEE Access 7:83771–83784 69. Xu X, Fu S, Cai Q, Tian W, Liu W, Dou W, Sun X, Liu AX (2018) Dynamic resource allocation for load balancing in fog environment. Wirel Commun Mobile Comput 2018 70. Fahs A, Pierre G (2019) Proximity-aware traffic routing in distributed fog computing platforms 71. Lee H, Kwon B, Kim S, Lee I, Lee S (2015) Theoretical-analysis-based distributed load balancing over dynamic overlay clustering. IEEE Trans Veh Technol 65(8):6532–6546 72. Fernando N, Loke SW, Rahayu W (2016) Computing with nearby mobile devices: a work sharing algorithm for mobile edge-clouds. IEEE Trans Cloud Comput (2016) 73. Lu H, Gu C, Luo F, Ding W, Liu X (2020) Optimization of lightweight task offloading strategy for mobile edge computing based on deep reinforcement learning. Future Gener Comput Syst 102:847–861 74. Fan Q, Ansari N (2018) Towards traffic load balancing in drone-assisted communications for IoT. IEEE Internet Things J 6(2):3633–3640 75. Han T, Li S, Zhong Y, Bai Z, Kwak K-S (2019) 5G software-defined heterogeneous networks with cooperation and partial connectivity. IEEE Access 7:72577–72590 76. Li C, Wang YaPing, Tang H, Zhang Y, Xin Y, Luo Y (2019) Flexible replica placement for enhancing the availability in edge computing environment. Comput Commun 146:1–14 77. Li C, Sun H, Chen Y, Luo Y (2019) Edge cloud resource expansion and shrinkage based on workload for minimizing the cost. Future Gener Comput Syst 101:327–340 78. Asif-Ur-Rahman M, Afsana F, Mahmud M, Shamim Kaiser M, Ahmed MR, Kaiwartya O, James-Taylor A (2018) Toward a heterogeneous mist, fog, and cloud-based framework for the internet of healthcare things. IEEE Internet of Things J 6(3):4049–4062 79. Bulkan U, Dagiuklas T, Iqbal M, Huq KMS, Al-Dulaimi A, Rodriguez J (2018) On the load balancing of edge computing resources for on-line video delivery. IEEE Access 6:73916–73927 80. Ramaswamy L, Liu L, Iyengar A (2007) Scalable delivery of dynamic content using a cooperative edge cache grid. IEEE Trans Knowl Data Eng 19(5):614–630 81. Wan J, Chen B, Wang S, Xia M, Li D, Liu C (2018) Fog computing for energy-aware load balancing and scheduling in smart factory. IEEE Trans Ind Inf 14(10):4548–4556

Balancing Exploration and Exploitation in Nature Inspired Computing Algorithm K. Praveen Kumar, Sangeetha Singarapu, Mounika Singarapu, and Swaroop Rakesh Karra

Abstract Nature is the one of the best inspiration for solving problems around us. Nature inspired computing algorithms are inspired from nature. These algorithms have inbuilt features of self learning, self motivation, co-ordination and collective behavior for solving a particular task. These nature inspired computing algorithms are specified by local search (exploitation) and global search (exploration). As per the effects of nature inspired computing algorithms, depends the searching capability. This paper focuses balance of local and global searching attribute of nature inspired computing algorithms. The people in a group are managed by leaders. By applying PSO algorithm, the best leader in the group is selected using efficient fitness functions. And this proposed fitness function is compared with the benchmark fitness function for providing better optimum results in the balancing of exploration and exploitation. Keywords Nature inspired computing algorithms · Swarm intelligence · Particle swarm optimization · Local search · Global search

K. P. Kumar (B) · S. Singarapu Department of IT, KITSW, Warangal, India e-mail: [email protected] S. Singarapu e-mail: [email protected] M. Singarapu Dept. of MS ITM, Campbellsville University, Louisville, KY, USA S. R. Karra MS IT, New Castle Wilmington University, New Castle, DE, USA © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_12

163

164

K. P. Kumar et al.

1 Introduction Nature is the great inspiration for solving complex problems in the universe. Nature ˙Inspired Computing (NIC) algorithms are inspired from nature for solving complex problems and produce nearest results at reasonable amount of time [1, 2]. The NIC algorithms having features of collective behaviour, self learning, self adapt nature are focused to solve NP hard problems. The nature inspired computing algorithms have feature of random probability i.e., algorithms randomly pick solutions for better optimum results [3–6]. Particle Swarm Optimization (PSO) is a popular nature inspired computing algorithm proposed by Dr. Eberhart and Dr. Kennedy in 1995. PSO is influenced by social behavior of birds and animals [7–10]. PSO is a well known nature inspired computing algorithm for the optimization of initializing random population of solutions. PSO updates component generations for searching best positions. Their particles manage their velocity of own positions and also positions of their neighbor particles to get optimum solutions.

2 Existing Work Paper [11] focused on particle swarm optimization about local search (exploitation) and global search (exploration) by applying master and slave method. The leader swarm acts, master and other swarms act and then the slaves. The slave swarms provide new positions along best fitness values to master swarms. The position of new particles is updated by master swarm for best position and manage updated positions of themselves (master swarms) and update slave swarms’ position. Article [12] proposed reduced energy conservation model of power system for socio-economic development by applying modified particle swarm optimization. Balancing of local search as well as global search using a new technique of multiswarm was proposed. In the article, ‘clan’ was introduced in place of swarms, and every group had a leader. Representing local best is the best solution that is represented by the leader of clan [13, 14]. To select global best, all these leaders can represent themselves for selection procedure. The clans update their positions, and manage their population risks for selection of local and best positions accordingly.

2.1 Particle Swarm Optimization PSO is initialized with randomly generated population of particles (initial swarm) and a random velocity is assigned to each particle that propagates the particle in search space toward optima over a number of iterations. Each particle has a memory,

Balancing Exploration and Exploitation in Nature Inspired Computing …

165

remembering best position attained by it in the past, which is called personal best position (Pbest). Each particle has its Pbest and the particle with the best value of fitness is called global best particle (Gbest). Suppose that the search space is D dimensional, the ith particle of the population can be represented by a D-dimensional vector (xi1, xi2, …, xiD)T. The velocity of this particle can be represented by another D-dimensional vector (Vi1, Vi2, …, ViD)T. The previously best visited position of ith particle is denoted by Pi and the best particle in the swarm is denoted by Pg. The update of the particle’s position is accomplished by the following two equations. Eq. (1) calculates a new velocity for each particle based on its previous velocity and Eq. (2) updates each particle’s position in search space.  k    k (t) − xid (t) + c2 r2 pgk (t) − xid (t) Vidk+1 = wVidk + c1r1 pid

(1)

k+1 k+1 k xid (t + 1) = xid (t) + vid (t + 1)

(2)

where k = iteration number, d = 1,2,3,…, D, i = 1,2,3,…, N, N = swarm size, w = inertia weight which controls the momentum of particle by weighing the contribution of previous velocity, c1 and c2 are positive constants called acceleration coefficients and r1 and r2 are random numbers uniformly distributed between [0,1]. PSO particle represents D-dimensional area as particle i (particle i1, particle i2, particle i3… particle iD), where D represents dimension and particle iD minimum dimension, maximum dimension is mini and max limitations of Dth dimension. Velocity of particle i is velocity i = (velocity i1, velocity i2, velocity i3…. Velocity iD). Particles and time t made manipulations as per the following relations. V elocit y i (time t + 1) = velocit y i (time t) + r value 1 con 1 (P pbp − par ticle i (time t))

(3)

+ r value 2 con 2 (Pgbp − par ticle i (time t)) Par ticle i (time t + 1) = par ticle i (time t) + velocit y i (time t)

(4)

In above equation, r value1 and r value2 are random values in the range 0 and 1. The con 1 and con 2 are the velocity constants that improve the movement of particles for a given iteration. The last best position of ith particle is noted as Ppbp and the best position of ith. particle is noted as Pgbp. V elocit y i (time t + 1) = iner tia w x velocit y i (time t) + r value1con1 (P pbp − par ticle i (time t)) = r value2con2 (Pgbp − par ticle i (time t))

(5)

166

K. P. Kumar et al.

In the above equation, inertia is added [15]. For a perfect balancing between the global and local explorations through inertia, minimum iterations are considered to achieve optimal solution. iner tia = iner tia max =

iner tia max − iner tia min × iterations iterations max

(6)

In above equation, inertiamax is denoted by initial weight, inertiamin is denoted by final weights, iterationsmax is max number of iterations considered in each iteration which is denoted as the number of iterations at present. In addition, Linearly decreasing inertia weight method in PSO (LPSO) is a random inertia weight factor for dynamic systems tracking [16]. Inertia weight factor noted for random vary as per the following equation. iner tia = 0.5 −

rand() 2

(7)

In above equation, rand() denotes equally allocated random numbers in between values 0 to 1 range. Value 1.494 must be maintained by acceleration coefficients. This equation refers to Random weight method in PSO (RPSO).

3 Proposed Work 3.1 Improved Multiple Particle Swarm Optimization (IMPSO) The main objective of Extensible Multiple Swarm Particle Swarm Optimization is the co-ordination behaviour of communication among the people in the group ‘clan’ (group of people) as in Fig. 1. The leaders of the groups search for optimum solutions. The swarm (group birds) is called as clan (group of people), and every clan (group of members) have solutions produced by the members of group. The best member of every clan is treated as the leader of that clan. A new leader of the clan is selected every generation due to the changing of positions of the members for meeting participation. The inertia function control exploration (global best) presents the algorithm. The main task of clan is to manage and control other members of the clan for movement and time control. Based on the new available position, every leader meet at one room (location) to find the best leader.

Balancing Exploration and Exploitation in Nature Inspired Computing …

167

Fig. 1 Swarm optimization of clan with individual members

PSO algorithm is used for balancing the position for exploration as well as to find exploitation. Based on PSO, the Optimization of Meeting Room Method (OMRM) is performed. ˙In this, every member in the clan is equal to every particle in swarm. The position as well as velocity is updated by PSO. The proposed algorithm’s stepwise performance, at first step shows to select the best leader (in a group of clan) who is treated as local best. In second step, overall best leader (in clan groups) is selected and is treated as global best. The optimization is determined by OMRM in the meeting room as in Fig. 2. By executing this algorithm, those who are selected as the overall best leaders, share their position with other leaders.  i W eight nl +

 i W eight gl − i W eight nl x rand() = iteration max

(timet + 1) = iweight nl x V elocit yinl (t) Leader Leader + r valuecon(Par ticleglobal − Par ticlelocal (timet ))

par ticleinl (timet + 1) = par ticleinl (timet + 1) + V elocit yinl (timet )

(8)

(9) (10)

In above equation, nl denotes the normal_leaders, gl represents the overall global leader, partile i states the position of the normal_leaders, V elocit yinl states the velocity of normal leaders, i W eight gl and i W eight nl denote inertia weight of global leader and normal leader respectively.

168

K. P. Kumar et al.

Fig. 2 Meeting room approach

Algorithm for Improved Multi-Particle Swarm Optimization See Fig. 3.

4 Results and Discusssion The performance of Improved Multiple Particle Swarm Optimization (IMPSO) is evaluated using benchmark functions compared with MPSO, Master–Slave SPO (MSPO) [11], and Multi-swarm Cooperative Particle Swarm optimizer (MCPSO) [17]. Parameters used are same for all these algorithms. The number of iterations used is 500 with 50 dimensions. For every experiment, 30 runs are performed. Testing with function is shown in Table 1 and parameter values in Table 2. Table 3 indicates the performance optimum results of particles for benchmark functions. EMPSO perform best results when compare to other algorithms. IMPSO has 6 swarms, every group consists of 10 particles and 6 particles in the room. IMPSO has less computation and shows better performance for optimum solutions. In Fig. 4a, by applying Sphere function, simulation results of IMPSO shows better performance when compared with other algorithms. In Fig. 4b, by applying Griewank function, simulation results of IMPSO shows better performance than that of MCPSOS, SPSO, MPSO, IMPSO [12].

Balancing Exploration and Exploitation in Nature Inspired Computing …

169

Fig. 3 Improved multi particle swarm optimization algorithm

Table 1 Testing with function Function name Sphere unimodal Griewank unimodal

Function D f1 = i=1 X2 D D f2 = i−1 −x21/4000 − i=1 cos

xi √ i

5 Conclusion The paper shows the importance of Swarm Intelligence algorithms for solving real-time problems. Swarm Intelligence algorithms are related to Nature Inspired Computing algorithms. The social group behavior of human being is proposed by applying swarm intelligence algorithm. The proposed algorithm Improved Multipurpose particle swarm optimization manages the local search (exploitation) as well as the global search (exploration) techniques. By applying benchmark functions, the results of the extensible multi purpose swarm optimization shows better performance

170

K. P. Kumar et al.

Table 2 Parameter values Algorithm

Parameter

Value

SPSO

W (inertial weight)

0.9–0.4

MCPSO

MSPO

IMPSO

Number of swarms

1

Swarm size (population)

50

c1, c2 (velocity constants)

1.5

W (inertial weight)

0.9–0.4

Number of swarms

1

Swarm size

50

c1, c2(velocity constants)

1.5

Wln

0.8–0.5

Wlg

0.9–0.7

c1,c2(velocity constants)

1.5

Number of clans

5

Clan size

10

Inertianl

0.8–0.9

Inertiagl

0.9–0.9

Constant1, constant2

1.5

Number of clans

6

Clan size

10

Table 3 Benchmark test functions’ results Fn

Swarm

Algorithm

Best

F1

50

SPSO

2.5457521

MCPSO

0.9854126

MPSO

0.0007845

F2

50

EMPSO

0.0006555

SPSO

0.0884741

MCPSO

0.0078414

MPSO

0.0000897

IMPSO

0.0000500

than the existing swarm optimization algorithms. The future work plans to implement hybrid nature inspired computing algorithms for optimum results.

Balancing Exploration and Exploitation in Nature Inspired Computing …

171

Fig. 4 a Convergence curve using sphere function, b convergence curve using Griewank unimodal function

References 1. Binitha S, Sathya SS (2012) A survey of bio inspired optimization algorithms. Int J Soft comput Eng 2(2):137–151 2. Dixit M, Upadhyay N, Silakari S (2015) An exhaustive survey on nature inspired optimization algorithms. Int J Softw Eng Appl 9(4):91–104 3. Slowik A, Kwasnicka H (2018) Nature inspired methods and their industry applications— swarm intelligence algorithms. IEEE Trans Ind Inform 14:1004–1015 4. Diao R, Shen Q (2015) Nature inspired feature selection meta-heuristics. Artif Intell Rev 44:311–340

172

K. P. Kumar et al.

5. Azrag MAK, Kadir TAA, Odili JB, Essam MHA (2017) A global African Buffalo optimization. Int J Softw Eng Comput Syst 3:138–145 6. Odili JB, Kahar MNM, Anwar S (2015) African Buffalo optimization: a swarm-intelligence technique. Procedia Comput Sci 76:443–448 7. Eberhart R, Kennedy J (1995) A new optimizer using particle swarm theory. In: Proceedings of the 6th ınternational symposium on micro machine and human science 1995, MHS 1995, pp 39–43 8. Kennedy J, Eberhart R (1995) Particle swarm optimization. In: Proceedings of IEEE ınternational conference on neural networks 1995, vol 4, pp 1942–1948 9. Zhang Y, Wang S, Ji G (2015) A comprehensive survey on particle swarm optimization algorithm and its applications. Math Probl Eng 10. Kameyama K (2009) Particle swarm optimization-a survey. IEICE Trans Inf Syst 92(7):1354– 1361 11. Niu B, Zhu Y, He X, Wu H (2007) MCPSO: a multi-swarm cooperative particle swarm optimizer. Appl Math Comput 185:1050–1062 12. Vijayakumar T, Vinothkanna R (2020) Efficient energy load distribution model using modified particle swarm optimization algorithm. J Artif Intell 2(4):226–231 13. Korani W, Mouhoub M (2021) Review on nature-inspired algorithms. In: Operations research forum, vol 2, no 3. Springer International Publishing 14. Kumar A, Nadeem M, Banka H (2022) Nature inspired optimization algorithms: a comprehensive overview. Evolving Syst 1–16 15. Shi Y, Eberhart R (1998) A modified particle swarm optimizer. In: 1998 IEEE ınternational conference on evolutionary computation proceedings. IEEE World Congress on Computational Intelligence, pp 69–73 16. Eberhart RC, Shi Y (2001) Tracking and optimizing dynamic systems with particle swarms. In: Proceedings of IEEE congress on evolutionary computation, pp 94–97. IEEE, Seoul 17. Shi Y, Eberhart RC (1999) Empirical study of particle swarm optimization. In: 1999 IEEE ınternational conference on evolutionary computation proceedings. IEEE World Congress on Computational Intelligence, vol 3, pp 1945–1950

Blockchain Based Secure, Efficient, and Scalable Platform for the Organ Donation Process of Healthcare Industry Keyur Parmar, Vadlapudi Devanand Kumar, Neduri Leela Prasanth, Pranoppal, Kasa Charan Teja, Shriniwas Patil, and Kaushal A. Shah

Abstract Organ transplantation is one of the most effective medical procedures to save lives. An individual’s organs can save up to nine lives. However, individuals refuse to donate organs because of lack of awareness and trust in the procedure, leading to the reduction in the number of organ donors. Individuals who wish to donate organs have to go through a complex administrative process, and sometimes these donated organs are managed by unauthorised individuals. To encourage individuals who wish to donate organs, we need a secure, efficient, and scalable platform. In this article, we present our perspective on the blockchain based organ donation management, in particular, for organ donation between organ donors and patients. The proposed platform uses the smart contract to automate the organ donation process and reduces the overall time of organ donation process. The proposed blockchainbased Organ Donation Platform (ODP) help patients in finding a matching donor efficiently. The ODP facilitates the process of organ donation by a decentralized network ensuring security, integrity, and transparency that eliminates the intermediaries. We comparatively evaluate the performance of the proposed ODP with the state-of-the-art literature. The proposed ODP is not only secure and scalable, but also efficient and reliable to find matching donor without revealing their identities. Keywords Blockchain · Smart contract · Ethereum · Security · Public key cryptography · Organ donation · Healthcare · Transparency

1 Introduction Organ transplantation is a medical process to remove a non-functional organ from the patient and then surgically replace it with a new organ taken from a healthy person. K. Parmar (B) · V. D. Kumar · N. L. Prasanth · Pranoppal · K. C. Teja · S. Patil S. V. National Institute of Technology, Surat, India e-mail: [email protected] K. A. Shah Pandit Deendayal Energy University, Gandhinagar, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_13

173

174

K. Parmar et al.

Fig. 1 Number of living and deceased organ donations from 2013–2017 [1]

Organ transplantation is one of the successful treatments for patients. Patients in need of organs often die due to a lack of information about donors in time. Organ donation process is categorised into living organ donation and deceased organ donation. In living organ donation, organs are retrieved from a healthy living person and transplanted into patients. In deceased organ donation, the functioning organs are retrieved from a deceased person and transplanted into patients. The donation trend for living donors and deceased donors between the years 2013–2017 is shown in Fig. 1. The trend of donating organs is more in the case of living donors. Nevertheless, due to a lack of public awareness, organ donations in the case of deceased donors is less compared to living donors. The lack of public awareness results in the acute shortage of organs that otherwise can be used to save lives. Therefore, there is need to create awareness about the organ donation. The Government of India organises organ donation programs such as National Organ Tissue Transplant Organization (NOTTO) [2] and Regional Organ Tissue Transplant Organization (ROTTO) [2]. Although the Government of India organises organ donation programs, individuals hesitate to donate organs due to the lack awareness about the of current organ donation programmes. Therefore, there is a need of transparent, secure, and a scalable platform for the organ donation process. In organ donation, the challenge is to find a matching donor. The number of patients die because of an unavailability of a donor [1, 3]. Another challenge is the lengthy paperwork that delays the whole process of organ transplant. A non-relative organ donor has to go through a number of procedures, such as, getting consent from the state government or local hospital. Moreover, there is a lack of infrastructure for organ donation which makes the whole process complex even if the donors are available. The above challenges are tackled by creating a decentralised system for the organ donation process. The proposed ODP will help individuals who wish to donate their organs and patients who are seeking organs.

Blockchain Based Secure, Efficient, and Scalable Platform …

175

In this article, we use the blockchain, a distributed ledger to store the data and to provide tamper resistance. Blockchain technology also guarantees security and transparency [4]. Therefore, a number of individuals will join the ODP who wish to donate organs. The proposed work uses the smart contract [5–7] with Ethereum Network [8, 9] to make the organ donation process automatic and fast. The smart contract reduces the time required to complete due formalities. The proposed work ensures transparency between donors and recipients (patients) and helps patients in finding the vital organ. The proposed work solves the challenge of finding organ donors and provides a choice for patients to choose organ donors. The ODP is trustworthy because all records are maintained using blockchain. The ODP eliminates the illegal smuggling of organs by restricting illegal access to ODP. The ODP will help to save lives by connecting the suitable donor to the right patient at the right time. In this article, our contributions are as follows: 1. We present a comprehensive literature review to identify the challenges in the conventional organ donation process. 2. We comparatively evaluate the strengths and weaknesses of the state-of-the-art proposals in the domain of organ donation. 3. We propose a blockchain based platform to facilitate patients to find a matching donor. The proposed blockchain based platform helps in mitigating the issues found in conventional organ donation process. 4. We implement the proposed blockchain based platform to demonstrate the working of the proposed ODP. 5. We evaluate the security and performance of the proposed blockchain based platform to highlight its strengths and weaknesses. The rest of the article is organized as follows: In Sect. 2, we discuss the related work. In Sect. 3, we discuss the preliminaries related to the proposed system. Section 4 presents the proposed ODP. Section 5 presents the experimental implementation of the proposed ODP. In Sect. 6, we conclude the article by emphasizing our contribution.

2 Related Work Organ donation is a well-researched topic that provides patients and organ donors a common platform. Chen et al. [10] identified an issue in a rule-based strategy of prioritising kidney donation in kidney donation programs where exchanges are initiated by Altruistic Donors (ADs). Authors developed a software-based decision support system. In addition, authors considered and compared two graph-based organ allocation algorithms, namely, MEU-Parallel and MEU-Sequential [10], to find the correct match. However, the authors work does not emphasize more on other kidneyallocation algorithms.

176

K. Parmar et al.

Prajapati et al. [11] implemented an online system to manage blood and organ transplant. The online system helps in searching for blood in case of emergency and in maintaining the records of blood and organ donors. Kadam et al. [12] discuss the drawbacks of manual blood donation and the organ donation process. Authors proposed an electronic donation system that manages the records of blood and organ donors and enables patients to monitor and find the correct donor. Rastogi and Tiwari [13] implemented an online web-based system for organ donation and management. An administrator manages the system. The administrator holds the rights and privileges to print the organs list. However, the system depends on the administrator which makes the system centralised and vulnerable to insider threats. Similarly, Ali et al. [14] implemented a web application with the support of a mobile application for a blood donation management system. The existing systems manage organ donations, but these systems require a connection between organ donors, patients, and central trusted third-parties. The system that uses the central trusted third-party such as hospital agents or government agents is vulnerable to insider and outsider threats and discloses the organ donor’s identity. If anonymity is not provided to organ donors and patients, the central third-party can contact patients to prioritise the patient to receive organs by asking for extra money. Dajim et al. [15] implemented a decentralised organ donation application. Authors used blockchain technology [16, 17] to distribute the information. The use of blockchain makes the system secure and tamper-proof. However, anonymity is not provided for organ donors and patients. Rajan et al. [18] proposed a system for organ donation and transplantation using blockchain technology. Authors used smart contracts to reduce the cost of transactions. However, the system does not allow anyone to check the availability of organs. The existing systems fail to provide security, transparency, a tamper-resistance, and do not protect patients from insider and outsider adversaries. Lakshminarayanan et al. [19] identified the dynamic update issue in blood management system. The authors proposed and implemented blockchain-based blood management system using hyperledger fabric framework to track the blood trail. The challenge of tracing thee blood trail is modelled as a supply-chain management issue. The proposed system provides transparency in blood donation process. However, the use of hyperledger fabric framework makes system permissioned. Therefore, no one can access system to get blood donation related information without getting permission from the network. Similarly, Quynh et al. [20] proposed blockchain-based innovative system to manage blood information. Authors proposed the system based on the architecture of hyperledger fabric with support of supply and demand of blood. Hawashin et al. [21] proposed a private Ethereum blockchain-based solution to enable organ donation and transplantation management. The use of private blockchain makes system decentralized, secure, traceable, auditable, private, and trustworthy. However, the private blockchain makes the organ donation and transplantation not visible to people outside the organization.

Blockchain Based Secure, Efficient, and Scalable Platform …

177

3 Preliminaries In this section, we briefly discuss the Blockchain [16, 17], Smart Contracts [5–7], and Ethereum [22].

3.1 Blockchain “A blockchain is a continuously growing list of records called blocks, which are linked and secured using cryptography [23].” The idea to timestamp [24] all transactions is to maintain the order of transactions. In a blockchain, the new blocks are appended securely using cryptographic hash functions. One of the fundamental features of the blockchain network is that it has no central authority. The blockchain network is shared among all the nodes and each node has a ledger or its own database, and the same copy is maintained by all nodes of the network. Therefore, any data present in the blockchain is secure, transparent, and tamper-proof [25].

3.2 Ethereum Ethereum [22] is a stable platform that can be used to codify and exchanges based on contracts. Ethereum is an open-source, blockchain-based platform for creating and running safe smart contracts and distributed applications (DApps) [26].

3.2.1

Infura

Infura provides the tools and infrastructure with stable access to Ethere-um and InterPlanetary File System (IPFS). Infura enables developers to test and deploy its scaled blockchain applications.

3.2.2

Metamask

Metamask is a browser extension that enables users to create decentralised applications (DApps) [26] without using the Ethereum network as an Ethereum node. Metamask is also serves as wallet that stores the Ethers (cryptocurrency) and allows users to send and receive Ethers via a decentralised application (DApp) [26]. Metamask allows us to link Infura, another Ethereum node, and runs smart contracts on the node.

178

3.2.3

K. Parmar et al.

Web3.js

Web3.js is a collection of libraries that enables an HTTP or IPC connection to communicate with a local or remote Ethereum node. The Ethereum blockchain is accessed via the web3 JavaScript library. Web3.js can perform operations such as accessing user accounts, submitting transactions, and communicating with smart contracts. We use Web3 Truffle-HDWallet-provider. The Truffle-HDWallet-provider provides a convenient network to link Ethereum thro-ugh infura.io. For example, the HDWallet provider adds some truffle-required features like event filtering and transaction signing that are unavailable with Infura.

3.3 Smart Contracts Smart contracts [5–7] are self-contained programmes that run on the blockchain network. Smart contracts are programmed to perform actions on blockchain as per the business logic and requirements [27]. The smart contract executes functions if certain code and business logic conditions are met or not met.

4 The Proposed Organ Donation Platform Using Blockchain Technology We propose an efficient, secure and scalable platform to facilitate donors and patients to manage organ donation. The platform using which both donor and patient will be able to manage all the operations such as login and signup. Donors can provide the required information of the organs which he or she wish to donate, patients can also find the donors by searching for the required organ. We present the whole process of the ODP as shown in Fig. 2. There are two entities in the system, namely, a donor and a recipient (patient). The entities are connected to

Fig. 2 Proposed organ donation platform using smart contracts with ethereum

Blockchain Based Secure, Efficient, and Scalable Platform …

179

the platform through front-end service where they can easily perform the operations such as login, signup, and search for organs. The back-end service executes the required operations requested through the front end, and all records are appended to the blockchain. The use of smart contracts [5–7] help in automating the organ donation process. The algorithm is used to find the available donors for the required organs. The proposed algorithm to retrieve all donors with the same organ as recipient is as shown in Algorithm 1. Algorithm 1 Algorithm to retrieve all donors with the same organ as recipient Result: Donor details Query database for all donors where recipient.organ= donor.organ; if results.length >0 then Show available donors; From donors list select a donor based on city; View donor details to contact; else Print(“No donors available”); end

Next, we present another algorithm to find the correct donor for the patient to receive an organ and make the process more efficient so the patient can perform the transplantation in the available time. The algorithm to match donor and recipient based on location is presented in Algorithm 2. Algorithm 2 Algorithm to match donor and recipient based on location Result: Match donor and recipient so that they can contact each other Function matchDonorRecipient(addr ess donor I d, addr ess r eci pient I d): public payable if match f ound f or both donor and r eci pient then retrieve donor info using donor address; retrieve recipient info using recipient address; end End Function

The proposed ODP provides transparency by adding donors on the blockchain that makes all the donors visible to anyone on the blockchain network with cryptoghraphically generated keys. The cryptographically generated keys are used to protect the privacy of the donors. The algorithm to add a donor to the blockchain is shown in Algorithm 3.

180

K. Parmar et al.

Algorithm 3 Algorithm to add donors to the blockchain Result: Add Donor to the Blockchain Each donor has the following attributes donor I d, hash, organ, blood_gr oup, match f ound, exist; Validate donor credentials; Call function Add_donor; Function Add_donor(donor I D, memor y hash, organ_name, bgr oup, f actor ): public checkdonorexist(donorID): if T r ue then Donor s[donor I D] ←− donor (donor I D, hash, bgr oup, organ_name, f alse, tr ue) donor array. push(donor I D); end End Function

5 Results and Analysis In this section, we discuss the implementation of the proposed ODP. We also discuss each functionality implemented in detail. The ODP consist of three layers as follows: 1. HTML, CSS, Javascript is used as a front end user interface for the organ donor and recipient to provide all the functionalities. 2. A node.js server is used to interact with the Ethereum blockchain network and provides the required data to Javascript Frontend. 3. All the respective organ donor and recipient data are stored in a decentralised network using Ethereum. The recipient logs in with the public key provided after registering. The credentials are to be provided by the user while signing up. The details of both donors and recipients are available in the blockchain ledger. Therefore, the proposed ODP can verify whether it is an existing user or not, if yes, then the system redirects the same to the login page otherwise, on the signup page. There are two separate portals for both donors and patients while registering on the proposed platform. Organ donors will be asked for his or her basic information and also the list of organs donors wish to donate after filling out a consent form. A patient will be able to register with his or her basic information and the organ that is required by the patient along with his contact information as a consent form. Patients verifies the matched donor in the dashboard provided when patients log in, and when there are no matches, the dashboard shows the message, “no matches found” and suggests the active donor’s list as shown in Figs. 3 and 4, respectively. In order to verify the security strength of the proposed platform, we compare the proposed platform with similar proposals in the domain of organ donation. The comparison is given in Table 1. The comparison is based on the well-known security services provided by existing proposals for organ donation. Based on the comparison of different organ donation proposals, in our opinion, the existing proposals only provides a few security services. However, in the proposed ODP, the confidentiality is achieved using PKC [28], the integrity of an information is verified by

Blockchain Based Secure, Efficient, and Scalable Platform …

181

Fig. 3 Dashboard when donor is matched with recipient

Fig. 4 Dashboard when no donor is matched with recipient Table 1 Comparison of different organ donation systems Authors A B C D Chen et al. [10] Prajapati et al. [11] Kadam et al. [12] Rastogi and Tiwari [13] Ali et al. [14] Dajim et al. [15] Ranjan et al. [18] Proposed Scheme

× ✓ × × × × × ✓

× × × × × × ✓ ✓

× × × × × ✓ ✓ ✓

× × × × × ✓ ✓ ✓

E

F

G

H

× × ✓ ✓ ✓ × × ✓

× × × × × × ✓ ✓

✓ × × × × × × ✓

✓ × ✓ ✓ × ✓ ✓ ✓

A awareness, B confidentiality and privacy, C decentralization, D tamper-resistance, E transparent, F automation, G algorithms to find match, H eliminate trusted third-party such as organ banks

182

K. Parmar et al.

comparing message digest of the information received along with encrypted information and with message digest of decrypted information. Privacy is achieved using an unique cryptographically generated key and it is unrelated to the identity of the users. Decentralization, tamper-resistance, and transparency are provided using blockchain technology [16, 17]. The protection from insider and outsider adversaries is achieved using timestamp. The platform is automated using smart contracts [5–7].

6 Conclusions and Future Works In this article, we propose a platform for organ donation using blockchain technology. To protect organ donors and patients from adversaries, we need a platform that provides transparency in the organ donation system. The ODP uses blockchain technology to make the central trusted third-party, e.g. hospital agent or government agent, more transparent and accountable while managing organ donation. The ODP is trustworthy and significantly reliable. In addition, the ODP protects organ donors and patients from insider and outsider adversaries. As a future work, there is a possibility to use the optimised algorithm that supports multiple cities and hospitals as an authority to verify the donors and patients. Acknowledgements This research was a part of the project “Design and Analysis of Secure and Efficient Smart Contracts Using Blockchain Technology”. It was partially supported by the SEED Money/Research Grant of the author, Dr. Keyur Parmar, Department of Computer Science and Engineering, S. V. National Institute of Technology (NIT), Surat, India.

References 1. Ahammed Mekkodathil MA, Sathian B, Rajesh E, Kumar RN, Simkhada P, van Teijlingen E (2019) Current scenario of organ donation and transplantation in Kerala India. Nepal J Epidemiol 9(2):759. https://doi.org/10.3126/nje.v9i2.24679 2. National organ tissue transplant organization. https://notto.gov.in/download-forms.htm. Accessed 03 Apr 2022 3. Nallusamy S et al (2018) Organ donation-current Indian scenario. J Pract Cardiovasc Sci 4(3):177–177. https://doi.org/10.4103/jpcs.jpcs_59_18 4. Abhishek B, Panjanathan R, Sarobin VR, Raja BE, Narendra M (2022) Data security in e-health monitoring system. Mater Today Proc 62:4620–4628. http://orcid.org/10.1016/j.matpr.2022. 03.079 5. Szabo N (1997) Formalizing and securing relationships on public networks. First Monday 2(9). https://doi.org/10.5210/fm.v2i9.548 6. Szabo N (1994) Smart contracts 7. Wang S, Ouyang L, Yuan Y, Ni X, Han X, Wang FY (2019) Blockchain-enabled smart contracts: architecture, applications, and future trends. IEEE Trans Syst Man Cybern Syst 49(11):2266– 2277. https://doi.org/10.1109/TSMC.2019.2895123 8. Buterin V (2022) Ethereum: a next-generation smart contract and decentralized application platform. https://ethereum.github.io/yellowpaper/paper.pdf. Accessed 30 Apr 2022

Blockchain Based Secure, Efficient, and Scalable Platform …

183

9. Wood G et al (2014) Ethereum: a secure decentralised generalised transaction ledger. Ethereum Proj Yellow Pap 151:1–34 10. Chen Y, Li Y, Kalbfleisch JD, Zhou Y, Leichtman A, Song PXK (2012) Graph-based optimization algorithm and software on kidney exchanges. Trans Biomed Eng 59(7):1985–1991. https://doi.org/10.1109/TBME.2012.2195663 11. Prajapati Mayur, Dungar Bhati YSHJ (2017) Online blood and organ transplant management system. Int J Eng Sci Res Technol 6(3):17–21. https://doi.org/10.5281/zenodo.266682 12. Nikhil Kadam N, Akshay Kaware VG (2020) Blood and organ donation system. Int J Sci Res Eng Dev 3(2):44–48 13. Rastogi S, Tiwari S (2019) Online organ donation management system. Int J Sci Res Eng Dev 5(10):445–448 14. Ali A, Jahan I, Islam A, Parvez S (2015) Blood donation management system. Am J Eng Res 4(6):123–136 15. Dajim LA, Al-Farras SA, Al-Shahrani BS, Al-Zuraib, AA, Merlin Mathew R (2019) Organ donation decentralized application using blockchain technology. In: Proceedings of the international conference on computer applications information security (ICCAIS), Riyadh, Saudi Arabia. IEEE, pp 1–4. https://doi.org/10.1109/CAIS.2019.8769459 16. Di Pierro M (2017) What is the blockchain? Comput Sci Eng 19(5):92–95. https://doi.org/10. 1109/MCSE.2017.3421554 17. Nofer M, Gomber P, Hinz O, Schiereck D (2017) Blockchain. Bus Inf Syst Eng 59(3):183–187. https://doi.org/10.1007/s12599-017-0467-3 18. Ranjan P, Srivastava S, Gupta V, Tapaswi S, Kumar N (2019) Decentralised and distributed system for organ/tissue donation and transplantation. In: Proceedings of the conference on information and communication technology, Allahabad, India. IEEE, pp 1–6. https://doi.org/ 10.1109/CICT48419.2019.9066225 19. Lakshminarayanan S, Kumar P, Dhanya N (2020) Implementation of blockchain-based blood donation framework. In: Proceedings of ICCIDS: international conference on computational intelligence in data science, Chennai, India. Computational intelligence in data science. Springer, pp 276–290. https://doi.org/10.1007/978-3-030-63467-4_22 20. Quynh NTT, Son HX, Le TH, Huy HND, Vo KH, Luong HH, Tuan KNH, Anh TD, DuongTrung N et al (2021) Toward a design of blood donation management by blockchain technologies. In: Proceedings of ICCSA: international conference on computational science and its applications. Lecture notes in computer science. Springer, Cagliari, Italy, pp 78–90. https:// doi.org/10.1007/978-3-030-87010-2_6 21. Hawashin D, Jayaraman R, Salah K, Yaqoob I, Simsekler MCE, Ellahham S (2022) Blockchainbased management for organ donation and transplantation. IEEE Access 10:59013–59025. https://doi.org/10.1109/ACCESS.2022.3180008 22. Buterin V (2022) On public and private blockchains. https://blog.ethereum.org/2015/08/07/ on-public-and-private-blockchains/. Accessed 28 Apr 2022 23. Vos J (2017) Blockchain and land administration: a happy marriage? Eur Property Law J 6(3):293–295. https://doi.org/10.1515/eplj-2017-0018 24. Haber S, Stornetta WS (1990) How to time-stamp a digital document. In: Proceedings of CRYPTO 90: advances in cryptology, conference on the theory and application of cryptography, Santa Barbara, USA. vol 537. Lecture notes in computer science. Springer, pp 437–455. https:// doi.org/10.1007/3-540-38424-3_32 25. Nakamoto S (2008) Bitcoin: a peer-to-peer electronic cash system. Decentralized Bus Rev 26. Metcalfe W et al (2020) Ethereum, smart contracts, dapps. Blockchain Crypt Currency 77. https://doi.org/10.1007/978-981-15-3376-1_5 27. Bashir I (2018) Mastering blockchain: distributed ledger technology, decentralization, and smart contracts explained, 2nd edn. Packt Publishing Ltd, ISBN: 978-1-78883-904-4 28. Al-Riyami SS, Paterson KG (2003) Certificateless public key cryptography. In: Proceddings of ASIACRYPT: advances in cryptology, international conference on the theory and application of cryptology and information security, Taipei, Taiwan. vol 2894. Springer, pp 452–473. https:// doi.org/10.1007/978-3-540-40061-5_29

Image Enhancement in Frequency Domain Fingerprint Detection and Matching Approach Suhasini S. Goilkar and Shashikant S. Goilkar

Abstract In digital image processing image enhancement techniques are used to improve the observation, perception and interpretability of the image information through human visual system. Image enhancement is implemented in spatial domain approach that operates directly to the pixels and frequency domain approach that operates through Fourier transform of an image. Image enhancement techniques in frequency domain are useful in different fields like early detection of physiological disorder, remote sensing, forensic science and biometric science. This implementation is done in frequency domain approach that means sharpening, smoothing and homomorphic filters are designed, implemented and image is enhanced to give the better input to the image processing automated techniques like biomedical, recognition and matching applications. In this research work, frequency domain Fourier transform techniques are designed and implemented for fingerprint detection and matching social applications. The frequency domain analysis performance is measured with performance measures which are peak signal to noise ratio and contrast to noise ratio with mean and variance. The fingerprint detection and matching progression technique is usually disintegrated into image pre-processing, matching and extraction. The designed, implemented and presented results of this research work will be very useful in social forensic science fingerprint matching different applications to improve the accuracy before applying to matching process. Keywords Fingerprint image · Frequency domain transforms · Image enhancement · Peak signal to noise ratio · Contrast to noise ratio

1 Introduction Now a days everywhere we are using digital techniques and to prefer digital techniques so many advantages are there as compared to analog techniques. This is clear and understandable that to prefer the image signal processing digitally for achieve the S. S. Goilkar (B) · S. S. Goilkar Finolex Academy of Management and Technology, Ratnagiri, Maharashtra, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_14

185

186

S. S. Goilkar and S. S. Goilkar

computer processing capabilities. Digital image processing is a very important tool to know and understand finer and finer details of any image. To process on the image digital signal with computer algorithms is nothing but the digital image processing. In digital image processing different techniques are used to perform different operations related to different applications. Image enhancement is the popular and important technique used for analysis and sharping of the different image features like contrasts, boundaries, edges of the digital image for the analysis and display purpose. The analysis and design of this approach is possible in time domain and frequency domain also. In frequency domain approach more emphasize of the image features is easily possible. From the input processing data, the intrinsic information contents will not increase in enhancement techniques, but the dynamic range definitely increased and because of that specific feature of the input image will be easily detected. In spatial domain image enhancement techniques from simple point processing to bit plane slicing total fourteen techniques are possible to use in the process of image enhancement, using these techniques possible to improve better and better human visual appearance of the original image for pre-processing analysis [1–3]. The highlighted image processing operations are gray level and contrast manipulation, edge sharpening and crispening, noise removal, magnification and interpolation etc. The image enhancement techniques are used to emphasize the silent features of the original data and simplify the processing task. In frequency domain image enhancement approach Fourier are used in transform and other frequency domain transforms are used for two dimensional images. Some transforms are used for the process to enhance the image visualization, feature selection, interest area measurements. Some transforms are used for image compression to reduce the data from original image with increased efficiency for the purpose of transmission and storage [4, 5]. In this paper, Image Enhancement in Frequency Domain Fingerprint Detection and Matching Approach is proposed and the comparative analysis is done with different cut off distance and filter order using PSNR and CNR parameters. The Section two highlights the important aspects of enhancement approach in frequency domain. In Section three adopted methodology is elaborated in detail. In section four designed and implemented filtering process steps are given. Finally, the simulated results are discussed in Section five and the entire work is concluded in Section six in Conclusion.

2 Frequency Domain Approach in Enhancement In frequency domain approach is the method in which it modifies the Fourier transform of an image. That means first required to compute the Fourier transform of the image. After that with this Fourier transform of the image required to multiply with filter transfer function. To get the modified image now need to compute inverse Fourier transform [6, 7]. In this operation the significant key parameters are the filter transfer functions like lowpass filters, high pass filters and Butterworth filters.

Image Enhancement in Frequency Domain Fingerprint Detection …

187

In image enhancement in frequency domain first required to compute the Fourier transform of a particular image which we want to enhance and just multiply it with filter function, convolution is not required and finally take the inverse transform for modified image. In frequency domain approach, it is instinctively simple to understood for image enhancement or idea of image blurring and deblurring to reduce high frequency components and for image sharpening to increase the magnitude of high frequency components. In spatial domain approach these operations are also effective with convolution using spatial filters. [8–10]. In the process of lowpass filtering it eliminates the image high frequency components to reduce the noise and blurring. In ideal low pass filtering possible to recall all low frequency components and to eliminate all high frequency components. But here in this computation process ideal low pass filter suffers with blurring and ringing effects and introduced large number if ripples or waves in the output image. In this the Butterworth filter in frequency domain is used for better results and smooth transitions in the computational process. The image captured from camera, or any sensor contains the light which is reflected from the object which we want to capture the image [11]. Generally, in the basic nature of the image is categorized in to two components that is amount of light incident on the object that is illumination to view the object and the amount of light reflected from the object that is reflection. In image enhancement homomorphic filters are used to enhance illumination and reflection components separately [12, 13]. In frequency domain image enhancement analysis low pass or smoothing filter, high pass or sharpening filter, homomorphic filter and colour image. In frequency domain using filters possible to sharp, smooth, deblur and restore the image. Basically, filtering process means the convolution of a function with filter function. Filtering process in frequency domain approach is concise below in Fig. 1. In the filtering process frequency domain approach, the significant points are to be understood that compute DFT transform of an image in frequency through filter and required to take inverse DFT to obtain filtered image [14, 15]. In frequency domain approach Fourier transform is applied for 2D images with different reasons Fig. 1 Filtering process in frequency domain approach

188

S. S. Goilkar and S. S. Goilkar

like to enhance the perceptibility of the image, selection of any part of an image, area of interest and measurement image compression for storage and transmission with increased efficiency. For the transmission and storage purpose fast Fourier transform is used for its efficient and powerful algorithm [16–18]. After image enhancement process the quality of the image is evaluated with two techniques or parameters that is qualitative and quantitative. In quantitative technique to check the performance psnr and variance values are calculated [19–21].

3 Adopted Methodology In this work fingerprint images are captured from fingerprint module and stored for further processing. For this fingerprint input stored image, the enhancement and filtering process in frequency domain is given in the block diagram shown below in Fig. 2. As shown in the above block diagram first step is required to do the pre-processing on the input fingerprint images like convert in gray image, resize if required and format the image. Now apply the frequency domain image enhancement techniques that are smoothing filtering, sharpening filtering and homomorphic filtering. In this implementation Butterworth and gaussian filtering techniques are implemented given in detail in the next session. At the end after simulation resulted images are stored and PSNR and CNR parameters are calculated on the basis of these parameters comparison is done. To calculate the peak signal to noise ratio matric alike signal to noise ratio having high values also shows the sign of more precise denoising.

Fig. 2 Enhancement and filtering process in frequency domain approach

Image Enhancement in Frequency Domain Fingerprint Detection …

189

4 Design and Implementation In the design and implementation of frequency domain algorithms the basic three parameters are considered that is in the Fourier transform edges and sharp changes mainly contribute with high frequency contents or low frequency contents. The smoothing is done with filtering high frequency contents and Butterworth, gaussian and ideal filters are used to enhance the fingerprint image. A. Smoothing frequency domain algorithm steps 1. 2. 3. 4.

Capture the fingerprint image using scanner or module and convert to gray. Resize and format. Obtain FFT and compute filter transfer function for three filters. Apply all three smoothing filters with different cut-off frequencies and filter orders. 5. Apply transfer function and do convolution operation. 6. Compute Inverse FFT and obtain the new enhanced image. 7. Calculate PSNR and CNR. B. Sharpening frequency domain algorithm steps 1. 2. 3. 4.

Capture the fingerprint image using scanner or module and convert to gray. Resize and format. Obtain FFT and compute filter transfer function for three filters. Apply all three sharpening filters with different cut-off frequencies and filter orders. 5. Apply transfer function and do convolution operation. 6. Compute Inverse FFT and obtain the new enhanced image. 7. Calculate PSNR and CNR. C. Homomorphic filtering frequency domain algorithm steps 1. Capture the fingerprint image using scanner or module and convert to gray. 2. Resize and format. 3. Obtain logarithm function as Fourier transform of low frequencies as illumination and high frequencies as reflection. 4. Determine the filter function which compress the gray level range and enhance the contrast. 5. Apply transfer function and do convolution operation. 6. Compute Inverse logarithm and obtain the new enhanced image. 7. Calculate PSNR and CNR.

190

S. S. Goilkar and S. S. Goilkar

5 Simulation Results and Discussion Finger print frequency domain image enhancement three techniques implemented and effectively completed. The image parameters like PSNR and CNR are extracted and the features are compared. In this implementation Butterworth filter, gaussian filter and ideal lowpass filters are used for fingerprint image enhancement. In these three-filter design filter order n = 4 and n = 8 is used and the cut-off distance D = 24 and D = 36 is used. The fingerprint enhanced images are shown below Fig. 3. The performance parameters PSNR and CNR are calculated for three filters with different filter orders and cut-off distance as shown in Table 1. For Parameter comparations the cutoff distance D = 24 and D = 36 is taken and the filter orders are n = 4 and n = 8 are used for all three filters. The Parameter comparation graphs of Butterworth filter, Gaussian filter and Ideal lowpass filter are shown below Fig. 4. In the comparation graph different cut off distance, filter order, different filters are taken to compare and show the comparation of psnr and cnr parameters.

6 Conclusion Image enhancement techniques are used to improve the image observation Image enhancement techniques in frequency domain are useful in different fields like early detection of physiological disorder, remote sensing, forensic science and biometric science. This implementation is done in frequency domain approach that means sharpening, smoothing and homomorphic filters are designed, implemented and image is enhanced to give the better input to the image processing automated techniques like biomedical, recognition and matching applications. The fingerprint detection and matching progression technique is usually disintegrated into image pre-processing, matching and extraction. The designed, implemented and presented results of this research work will be very useful in social forensic science fingerprint matching different applications to improve the accuracy before applying to matching process.

Image Enhancement in Frequency Domain Fingerprint Detection …

a) Finger print image

b) With Butterworth filter

a) Finger print image

b) With Butterworth filter

a) Finger print image

b) With Gaussian filter

Fig. 3 Finger print images and enhanced with three different filters

191

192

S. S. Goilkar and S. S. Goilkar

a) Finger print image

b) With Ideal filter

a) Finger print image

b) With Ideal filter

Fig. 3 (continued) Table 1 Parameter comparations with different filters Filter type

Cut off distance

Filter order

PSNR

CNR

Butterworth filter

D = 24

n=4

43.51

0.0067

n=8

43.49

0.0064

D = 36

n=4

43.72

0.0066

n=8

43.75

0.0063

D = 24

n=4

43.69

0.0067

n=8

43.73

0.0065

n=4

44.01

0.0064

n=8

44.09

0.0064

D = 24

n=4

43.61

0.0058

n=8

43.62

0.0056

D = 36

n=4

43.82

0.0056

n=8

43.85

0.0055

Gaussian filter

D = 36 Ideal low pass

Image Enhancement in Frequency Domain Fingerprint Detection …

193

Fig. 4 Parameter comparation graphs of Butterworth filter, Gaussian filter and Ideal lowpass filter

194

S. S. Goilkar and S. S. Goilkar

Acknowledgements The authors acknowledge the financial support from University of Mumbai, Maharashtra, India through the Minor Research Grant for this Project.

References 1. Daluz HM (2015) Fundamentals of fingerprint analysis. Taylor & Francis Group 2. Ali MMH, Mahale VH, Yannawar P, Gaikwad AT (2016) Overview of fingerprint recognition system. In: International conference on electrical electronics and optimization techniques ICEEOT, pp 1334–1338 3. Adam EEB (2021) Evaluation of fingerprint liveness detection by machine learning approach-a systematic view. J ISMAC 3(01):16–30 4. Ahmed ZJ, George LE (2017) Fingerprints recognition using the local energy distribution over haar wavelet subbands. Int J Sci Res 6(9) 5. Wenchao W (2012) A Fingerprint identification algorithm based on wavelet transformation characteristic coefficient. ICSAI 2–4 6. Tang T (2012) Fingerprint recognition using wavelet domain features. Int Conf Nat Comput 531–534 7. Dhannoon BN (2017) Fingerprint recognition by using iterative closest point. 7(4) 8. Dautov ÇP (2018) Wavelet transform and signal denoising using wavelet method. Signal Process Commun Appl Conf 1–4 9. Kwaochai A, Pongyupinpanich S, Areekul P, San-Um W (2017) An application program of fingerprint detection using wavelet transform for authentication. In: Management and innovation technology international conference (MITicon), pp MIT-217–MIT-220 10. Tewari K, Kalakoti RL (2014) Fingerprint recognition and feature extraction using transform domain techniques. In: International conference on advances in communication and computing technologies (ICACACT ), pp 1–5 11. Krishnasamy P, Kriegman D, Belongie S (2011) Wet fingerprint recognition: challenges and opportunities. In: International joint conference on biometrics (IJCB), pp 1–7 12. Hong L, Wan Y, Jain A (1998) Fingerprint image enhancement: algorithm and performance evaluation. IEEE Trans Pattern Anal Mach Intell 20:777–789 13. Chikkerur S, Cartwright A, Govindaraju V (2007) Fingerprint enhancement using STFT analysis. Pattern Recogn 40:198–211 14. Maltoni D, Maio D, Jain A, Prabhakar S (2009) Handbook of fingerprint recognition, 2nd edn. Springer, London 15. Lin C, Kumar A (2018) Matching contactless and contact-based conventional fingerprint images for biometrics identification. TIP 27(4):2008–2021 16. Dyre S, Sumathi CP (2016) A survey on various approaches to fingerprint matching for personal verification and identification. Int J Comput Sci Eng Surv (IJCSES) 7(4):1–7 17. Dyre S, Sumathi CP (2014) Hybrid approach to enhancing fingerprint images using filters in the frequency domain. In: IEEE international conference on computational intelligence and computing research (ICCIC), pp 1–6 18. He Z, Zhao X, Zhang (2015) Low-quality fingerprint recognition using a limited ellipse-bandbased matching method. J Opt Soc Am A 32(6):1171–1179 19. Chandra E, Kanagalakshmi K (2011) Noise elimination in fingerprint images using median filter. Int J Adv Netw Appl 02(06):950–955 20. Thai D, Huckemann S, Gottschlich (2015) Filter design and performance evaluation for fingerprint image segmentation. arXiv:1501.02113 [cs.CV] 21. Turroni F, Maltoni D, Cappelli R et al (2011) Improving fingerprint orientation extraction. IEEE Trans Inf Forensics Sec 6(3):1002–1013

Developing Machine Learning Based Mobile App for Agriculture Application R. Dhivya and N. Shanmugapriya

Abstract With the help of machine learning algorithms including KNN, SVM and LDA; it is possible to determine which crops are to be grown, when the soil needs more water and fertilizers, and what pests are present in the crops. The proposed system is designed to collect the current soil conditions and to calculate soil nutrients by analyzing the current soil conditions. In the proposed model, SVM offers 100% precision and LDA offers 95% accuracy in soil prediction. IoT camera sensor modules will assist farmers in determining whether their crops have been infected with pests so that they can take appropriate measures. In the application, the farmer can get alert notifications as well as other relevant information about crops such as soil type, crop types, nitrogen, potassium, and phosphorus, based on the soil and weather conditions. Farmers can also determine which crops to plant based on the soil and weather conditions. This empowers the farmer to take appropriate action for minimizing crop loss and maximizing crop yield. Keywords Machine learning · IOT · Soil condition · Mobile application · Camera sensor module

1 Introduction In order for a crop to be successful, farmers need to pay attention to a wide range of factors, including the quality of the soil, the amount of water that is in the soil, the temperature of the soil, and the amount of fertilizer that is used. At various points in their development and depending on the type of soil they are growing in, R. Dhivya (B) Department of Computer Science, Dr. SNS Rajalakshmi College of Arts and Science, 49, Coimbatore, Tamil Nadu, India e-mail: [email protected] N. Shanmugapriya Department of Computer Applications (PG), Dr. SNS Rajalakshmi College of Arts and Science, 49, Coimbatore, Tamil Nadu, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_15

195

196

R. Dhivya and N. Shanmugapriya

plants require varying concentrations and proportions of specific nutrients. Before the farmer can start to see any returns on his labor, he has to put in a significant amount of time and energy. Cut down on their workload by coming up with an answer that satisfies their requirements. The amount of water that is required is determined by the current moisture content of the soil as well as the climate outside. The amount of insecticide and fertilizer that should be applied is determined, in part, by the weather that is currently taking place. Our main goal is to provide a system that can detect and recommend the requirements to increase crop earnings. As a result, the farmer avoids losses thanks to our system’s earlier pest detection. By analyzing the climate and soil conditions, it also proposes what crop should be produced. in order to increase profits, the crop’s market demand. Now, Sect. 2 of the paper below discusses how relevant the suggested idea is. The part III that follows Sect. 2 discusses the literature review that was conducted. The flow of the suggested solution is presented in section IV after which each module is thoroughly described. Our suggested solution’s results are finally put to rest.

2 Relevance To date, farmers have only added fertilizer, sprayed pesticides, and increased soil moisture based on their own experience, expertise, and observations of the existing state of the land. Instead of relying on the present weather and soil conditions, farmers themselves decide how much water, insecticides, and fertilizer should be applied, how often, and how much of each. To effectively identify this, farmers generally lack the experience and knowledge necessary. A lot of time passes before soil testing is done. As a result of the fertilizers being added in accordance with the results of the soil testing, and because the present soil conditions may differ greatly, this may result in soil degradation. Additionally, pests are not found in agricultural produce at the proper period, which has a negative and destructive effect. Pests should be watched out for in crops, and appropriate action should be taken to address problems. Which crops should be grown cannot be determined with accuracy by farmers. The type of crop that has to be grown is chosen without taking present soil conditions, weather, or market demand into account. The issues highlighted above will be resolved by our suggested solution, which will also assist farmers in ensuring a better crop.

3 Related Work There is a list of suitable crops that has been provided by Jain [1], and it takes into account both the yield and the market prices. In addition to that, there is an algorithm for crop sequencing that can be used to rank the crops that are chosen. The CSM crop selection algorithm [2] is based on a variety of crop classifications, such as seasonal, year-round, short-term, and long-term plantations. These crop classifications are

Developing Machine Learning Based Mobile App for Agriculture …

197

used to determine which crops should be planted. The CSM algorithm takes into account a variety of factors, including the climate, the composition of the soil, the amount of available water, and the type of crop being grown. The author of [3] recommends harvesting crops through the use of a random forest methodology. The following are some of the criteria that are taken into consideration: weather, soil quality, moisture content, the due factor, irrigation, and yield form. An ARIMA model, a neural network model, and a multilinear regression model are incorporated into a proposed multi-model for price prediction that is presented in [4]. This model is intended to estimate crop market prices. The author of reference [5] presents a model for predicting prices. Methods of image processing such as masking and segmentation are described in [6], which was written by its author. The clustering method uses the relevant parts of the selected images that were left over after segmentation. In this instance, the image of the pest is obtained by first obtaining an image of the grouped plants and then subtracting the image of the pest from that image. In [7], the identification of pests is accomplished through the use of a few different image processing techniques. The SVM’s training was done with the help of color features. When using an SVM, it is much simpler to differentiate between the pixels representing the leaves and the insects. Agriculture continues to be the primary source of income for more than half of the Indian population in recent years. About 18% of the nation’s GDP is derived from agriculture, making it one of the most important sectors of the economy. All emerging technologies can boost crop yields by maintaining ideal growing conditions. Through an Android application, the internet of things is used to automate agricultural machinery in accordance with the article’s suggested approach. Through machine automation, all work can be performed without human intervention. It utilizes a beagle bone air processor, wasp mote sensors, and numerous other IoTenabled hardware components to automate agricultural fields. By using this method, agricultural fields could become more productive, and farmers’ fatigue could be reduced significantly [8]. It will be necessary for the world population to increase output in each industry, especially agriculture, in order to meet the growing demand for food. There will, however, be times when demand and supply diverge. It is challenging to increase agricultural productivity through management of capital, people, and resources. Smart agriculture is a better solution for maximizing food production, resource management, and labor. This research presents forecasting analysis, Internet of Things (IoT) devices with cloud management, and security units for multi-culture agriculture while taking previous farming experience into consideration. Statistical and quantitative methods can be used to revolutionize the agriculture system more effectively by integrating contemporary technologies and traditional farming practices. A green field also includes irrigation, plant diseases, crop phases, and drone activation via IoT. Sensors are triggered for a variety of purposes in the Internet of Things. Modern agriculture will be developed using cutting-edge IoT tools and ideas in this research. A systematic evaluation provides an overview of current and upcoming trends in the agriculture sector. There are numerous analysis sections in this study that provide a broad framework for the model presented. We have thus been able to provide a

198

R. Dhivya and N. Shanmugapriya

number of advantages to our integrated units, as we have previously stated in our article regarding smart agriculture units with IoT modules. There are also some restrictions associated with this restricted paradigm for platforms and security. Most IoT-based agricultural devices consider the number of difficulties and constraints. Devices that are connected to the Internet of Things aim for cost-effectiveness by reducing hardware and software costs without compromising their output precision. Imported gadgets ignore the compromise of reduced component costs. Additionally, standardizing the data format for the process will improve device consistency and execution speed. Initially, there is a process barrier for engaged farmers to buy products or services when the system is being enhanced. As a result of their interconnection via a web server, the proposed integrated system will be complex [9]. An important technique for improving accuracy and performance in the IoT industry is the heterogeneity property. Deep learning analysis can be used to increase the productivity of IoT-enabled smart agriculture by using a large number of characteristics or data.

4 Proposed Work The use of an Android application comes highly recommended as a method of operation. The capabilities of the application can be put to use in order to assist farmers in making decisions regarding which crops to cultivate, how much of each crop can be expected from a given yield, and how much money each crop will fetch on the open market. The farmer is informed, for one thing, of the current moisture content of the soil and is made aware of the presence of a pest as a result of using this tool. The system consists of NPK pills, a Moisture module, and an IoT camera module in order to ascertain the NPK values of the soil, ascertain the moisture content of the soil, and take photographs of pests. This strategy seeks to better the lives of farmers by addressing the problems they face and offering solutions to those problems.

5 Block Diagram The represented diagram depicts the block diagram of our system (Fig. 1).

6 Architecture Description The architecture of our solution is depicted in Fig. 1, which is a diagram. After a certain amount of time has passed, the IoT camera module and the moisture sensor will both send their data to Firebase. The farmer is responsible for inputting the NPK values into the application based on the readings from the pill chart. The hosted API

Developing Machine Learning Based Mobile App for Agriculture …

199

Fig. 1 Block diagram

for pest detection is called by the Raspberry Pi module that is connected to the Internet of Things camera. Notifications are dispatched in the event that a pest is identified through the mobile application used by farmers. The farmer initiates the API request for the other modules of the application by selecting one of the features available in the application. These modules include crop selection, yield prediction, and market price prediction. In order to acquire the required data, hosted APIs make use of two APIs: the weather API and firebase (weather data required for yield prediction and crop selection module). The desired outcomes are achieved by training models on the available input data. The results will be sent back to the application by the API. In the steps that follow, the results will be shown to the farmer on the screen of the application.

7 Live Field Status Real-time information on soil moisture is fetched by this module, which also alerts the farmer when to water the plants. A predetermined threshold value for soil moisture was used to compare the soil’s moisture level to. A notification detailing the situation will be sent to the farmer’s cell phone if the scale falls below the threshold. To do this, a soil moisture sensor called NodeMCU (ESP8266) is employed, continuously recording the field’s real-time status. The processing of this data then proceeds once it is pushed to the database. It is Firebase’s Realtime Database that is used for this purpose. You may create top-notch apps with the help of Google’s mobile platform, Firebase. A No-SQL database housed in the cloud called the Firebase

200

R. Dhivya and N. Shanmugapriya

Fig. 2 Dataset repesentation

Realtime Database enables real-time data syncing and storing between users. FirebaseESP8266.h is the ESP8266 library that is utilized for the transmission in NodeMCU.

8 Crop Detection and Market Prediction As part of this session, you will advise the farmer on the best crops to plant in order to increase their yield and profit.

8.1 Data Gathering The STCR Research Crop-wise Recommendations determined the necessary NPK values for a wide variety of crops by conducting research that was specific to each type of soil [10]. Figure 2 provides a visual representation of the dataset used for crop prediction. It includes information about the type of soil, the variety of crop, and the season, in addition to information about the amount of NPK that is currently available and that is required. On the Open Government Data (OGD) Platform in India, a dataset containing crop yield information was available. [11] Each of the following factors—crop year, season, area, and production—is broken out into its own column. Data on minimum and maximum temperatures as well as rainfall were supplied by Skymet’s weather services [12]. The data on market prices incorporate information on crop and market prices [13].

8.2 Data Pre-processing Using OpenCage Geocoder, Data Science Toolkit, and a climatic station that is located in the area, the crop yield data from an Indian district was combined

Developing Machine Learning Based Mobile App for Agriculture …

201

Fig. 3 Clustering

with climate parameters such as minimum and maximum temperatures and rainfall. Normalization of the datasets was necessary before they could be used as input for the training models [14–20]. The following are the prerequisite actions to take:

8.2.1

Clustering

Clusters are formed by crops that are grown in conditions that are comparable in terms of the soil and the climate. Figure 3 shows the methods of KNN, SVM, and LDA that are utilised in the process of accomplishing this objective. For instance, when establishing clusters, factors such as rainfall, temperature, and the type of soil are all taken into consideration. The model will generate a crop cluster that is appropriate for a given new observation by taking into account the current conditions of the soil and the weather.

8.2.2

Yield Prediction

The yield of the crop is predicted with the help of a decision tree regression model in this section. It is possible to prevent over pruning by limiting tree cuttings to a depth of no more than one quarter of an inch. As input features, we make use of the location (with a single Hot encoded), temperature, and precipitation.

8.2.3

Price Prediction

It is possible to determine the market price of the appropriate crops by employing a model that uses time series analysis. The Autoregressive Integrated Moving Average Model is utilised in order to accomplish this objective (ARIMA model). We determined whether or not the data were stationary by using the ADF test. One separate model will be trained for each crop and market in order to make a prediction about the possible selling price at that particular market location. The farmers can then use

202

R. Dhivya and N. Shanmugapriya

Fig. 4 Crop prediction using ML

this formula as a guide to assist them in selecting the most profitable crops to grow on their land in order to maximise their income.

9 Crop Detection 9.1 Image Acquisition Using color cameras, various images of the soil samples that need to be categorized are taken and delivered as input to the system. Each type of soil’s characteristics are gathered and kept in its own database. The final part of this process involves the detection of soil and crops using this database. The image’s edges and contrast need to be improved for greater accuracy and clarity. The image’s color maps are utilized to improve the contrast and edges. The crop prediction utilizing machine learning approaches is shown in Fig. 4.

9.2 Image Pre-processing There are some inaccuracies in the image we first obtained. The results of the analysis are significantly influenced by the image’s quality because it affects both the detection of the objects under investigation and the precision of the measurements that follow. To get an image free of errors, pre-processing steps are used. Since it involves improving the image by boosting contrast and correcting faults to produce a higher-quality image for use in later stages, this process is often referred to as “image augmentation.” Before continuing with the processing, the image must be cleaned up of any flaws like noise or artifacts like scratches, lap tracks, comet tails, etc. In order to remove the noise and artifacts from the image, a filter known as the Smoothing filter is used. Low pass filters and high pass filters are the two different categories of filters. A low pass filter is a smoothing filter. It is applied to digital images to reduce high spatial frequency noise. A moving window operator is used in smoothing filters to change the value of each individual pixel in the image by some function of a nearby

Developing Machine Learning Based Mobile App for Agriculture …

203

region of pixels. To affect every pixel in the image, the operator glides over it. The smoothing filter gradually improves the image by erasing the defects as a result of numerous rounds.

9.3 Feature Extraction Following the k-means stage, in which the image is broken down into its individual components, the next stage is the feature extraction stage. This is the first step in the process that needs to be completed. During this phase, both the crop detection and the soil type classification processes are finished. To determine the type of soil present, scientists look at a wide variety of characteristics, such as the texture, colour, intensity, saturation, and hue of the sample. Feature extraction is accomplished with the help of the Gabor Filter. The Gabor Filter is a type of linear filter that analyses images in order to locate and highlight edges. Because of the similarities between the human visual system and the Gabor filter, it has been discovered that the frequency and orientation representations of the Gabor filter are particularly suitable for representing and differentiating textures. Utilizing a collection of frequency and orientation representations can be helpful in extracting the most information possible from an image. In addition, the Gabor filter can be used to obtain information such as the mean, the standard deviation, and the entropy. It is necessary to extract the colour of the soil, as it is the most important component. This distinction can be made with the help of colour moments, which is a metric that compares the colour characteristics of various images to one another. By comparing the values of similarity between photos to the values of images stored in an image database, it is possible to carry out activities such as image retrieval using the values of similarity between photos.

9.4 SVM Classification The method known as Support Vector Machine (SVM) is utilised in the process of soil classification. It has found applications in a wide range of fields, including bioinformatics, recognition of text and images, as well as other areas. Because of how easy it is to use, SVM is a popular choice for an algorithm. This algorithm is being considered a potential replacement for the neural network algorithm. This is a description of how the process works, as it follows: An SVM training algorithm generates a model that sorts newly discovered examples into either of two categories. This classifier is a non-probabilistic binary linear classifier. Given a collection of training examples that have been labelled as belonging to one of two categories, it performs this action. Within an SVM model, there ought to be as much breathing room as is humanly possible between the examples of the various categories. After that, those newly created instances are mapped according to which side of the divide they are located on. SVM is helpful for finding the margin with the largest difference

204

R. Dhivya and N. Shanmugapriya

and performing separate plane analysis. This helps in determining whether or not the data points have sufficient support. As a consequence of this, the mapping looks like this: In this particular scenario, X is an instance of a class, and Y is a label for the category to which it belongs. This algorithmic approach is based on distinguishing characteristics analysis and looks at predicted error minimization. To improve the training process, this technique considered empirical risk. In this case, structural analysis is used to calculate risk in order to reduce generalization error. Based on a review of the error margin under the class deviation heading, the closest training patterns are identified. This model is similarly built on the polynomial kernel representation in order to efficiently learn the elements and achieve greater accuracy.

9.5 Segmentation The image is enhanced prior to the application of the segmentation technique by using image pre-processing techniques. K-means Image segmentation into its component parts is accomplished through the process of clustering. Partition clustering is a technique that is used to maximise a set of clustering criteria by dividing a given data set into disjoint subgroups. This technique is used as part of a partition clustering method. The clustering error criterion is the one that is utilised the vast majority of the time. This criterion calculates the squared distance that each point has from the centre of the cluster, and then adds up these distances for each point in the data set. The assignment of each pixel in the image to a cluster begins at the cluster’s centre and continues throughout the image. The centre of the cluster is located by taking the average of all of the pixels. The next step in this cycle entails connecting each point in a particular data set to the centroid that is located closest to it. K-means The algorithm is as follows: (1) K, a group of points ranging from × 1 to × n (2) Disperse the centroids C1………Ck throughout the universe. (3) Repeat till convergence. J = 1 for each cluster, K; The new centroid Cj from the previous step equals the average of all the points xi attributed to cluster J. (4) Stop if none of the cluster assignments change.

10 Implementation and Result Analysis 10.1 Home Screen Along with Application Menu With the help of this screen, the farmer will easily comprehend the app. Temperature, humidity, wetness, crop kind, soil type, nitrogen, potassium, and phosphorus content,

Developing Machine Learning Based Mobile App for Agriculture …

205

Fig. 5 Home screen

as well as the name of the fertilizer, are all displayed by the programme as illustrated in Fig. 5.

10.2 Confusion Matrix A confusion matrix is used to assess the classification model’s effectiveness. Measuring the effectiveness of our models by accuracy when we have uneven data can be deceptive. Classification models with binary or multiclass output are evaluated using performance metrics. Four of these pairings are listed in a table. There are two things to notice in the image up there. (a) Predicted values: Values that the model predicts (b) Actual Value: Datasets that include actual values. We will utilize binary classification in this instance to comprehend the model. Positive points belong in the positive class, whereas negative points belong in the negative class. Rate is a measuring factor in a confusion matrix. The other 4 categories are TPR, FPR, TNR, and FNR. Figure 6 displays the precision and accuracy of a machine learning system based on a confusion matrix.

Fig. 6 Precision and accuracy analysis based on KNN, SVM, LDA

206

R. Dhivya and N. Shanmugapriya

i.

Values that are both actual and expected to be positive are known as True Positive (TP) values. ii. Values that are anticipated to be positive but are actually negative are known as false positives (FP). iii. False negatives are values that are positive when they should be negative (FN). iv. Values that are both actual and expected to be negative are referred to as True Negative (TN) values.

10.3 ML Model Comparison Agriculture is viewed as a vital business around the world where there are several challenges in calculating crops based on environmental conditions. This has become an issue for developing nations. Utilizing the most recent technologies, many organizations are reducing manual labor by using mechanical technology and IOT-based services. When removing physical labor, these strategies are especially useful, but not when generating predictions. In this study, crop yield is forecasted based on soil and temperature data utilizing the most modern ML technology such as KNN, SVM, and LDA. A dataset is built with various soil conditions as features and labels in order to forecast which type of each label will be connected to a certain crop. The sort of crop most suited for the existing conditions will be determined by the user’s input of soil characteristics into the prediction procedure. Application assists in choosing the highest-yielding crops for hector. The examination of precision and accuracy based on the classification algorithms KNN, SVM, and LDA is shown in Fig. 7.

Fig. 7 Precision and accuracy analysis based on KNN, SVM, LDA

Developing Machine Learning Based Mobile App for Agriculture …

207

10.4 Notification and Information This module will help farmers comprehend the soil conditions on their holdings. A notification will be sent based on the soil type and crop type of a soil. The app will also display the soil’s temperature, moisture level, and humidity along with Nitrogen, Potassium and Phosphorous which is additional incorporated in this proposed model. The information on this page will assist other modules in carrying out their respective tasks. Figure 8 displays the suggested crop depending on the farmer’s input. Fig. 8 Precision and accuracy analysis

208

R. Dhivya and N. Shanmugapriya

11 Conclusion The proposed system offers the farmer a variety of modules in an attempt to address many of their problems. The technique offers soil prediction and crop prediction and helps to decrease losses in the agri field. The module lets the farmer know which crops are likely to thrive in his region based on soil and climate. With this module, farmers can sell their goods at higher prices. The smartphone application will allow us to learn everything about the farmers and their fields. In the suggested model, SVM provides 100% precision while LDA provides 95% accuracy for predicting soil. Using the machine learning technology, the system takes into account both the current soil conditions and climate in order to make better recommendations.

References 1. Jain N, Kumar A, Garud S, Pradhan V, Kulkarni P (2017) Crop selection method based on various environmental factors using machine learning. Int Res J Eng Technol (IRJET) 4(02) 2. Shyamalaprasanna A, Velnath R, Dhivya KT, Aishwarya S, Saravana G, Srimathi R (2021) Monitoring and controlling of industrial sewage outlet using IoT. In: 2021 International conference on advancements in electrical, electronics, communication, computing and automation (ICAECA), 2021, pp 1–5. https://doi.org/10.1109/ICAECA52838.2021.9675638 3. Karthikeyan R, Gowthami M, Abhishhek A, Karthikeyan P (2018) Implementation of effective crop selection by using the random forest algorithm. Int J Eng Technol 7(3):34. https://doi.org/ 10.14419/ijet.v7i3.34.19209 4. Ravanan V, Subasri R, Vimal Kumar MG, Dhivya KT, Kumar PS, Roobini K (2021) Next generation smart garbage level indication and monitoring system using IoT. In: 2021 Smart technologies, communication and robotics (STCR), 2021, pp 1–4. https://doi.org/10.1109/STC R51658.2021.9588961 5. Kaur M, Gulati H, Kundra H (2014) Data mining in agriculture on crop price prediction: techniques and applications. Int J Comput Appl 99(12):0975–8887 6. Priya P, Muthaiah U, Balamurugan M (2018) Predicting Yield of the crop using machine learning algorithm. Int J Eng Sci Res Technol. ISSN: 2277-9655; Krishnan M, Jabert G (2013) Pest control in agricultural plantations using image processing. IOSR J Electron Commun Eng (IOSR-JECE) 6(4):68–74. e-ISSN: 2278-2834,p-ISSN: 2278-8735 7. Rajan P, Radhakrishnan B, Padma Suresh L, Detection and classification of pests from crop images using support vector machine 8. Bashar A (2019) Agricultural machine automation using IoT through android. J Electr Eng Autom 1(2):83–92 9. Suma V (2021) Internet-of-Things (IoT) based smart agriculture in India-an overview. J ISMAC 3(01):1–15 10. STCR Research Crop Wise Recommendations. http://www.iiss.nic.in/downloads/stcr%20C rop%20wise%20Recommendations.pdf 11. Open Government Data OGD Platform India. https://data.gov.in/ 12. Skymet Weather Services. https://www.skymetweather.com/ 13. Market price data. https://agmarknet.gov.in/ 14. ImageNet dataset. http://www.image-net.org/ 15. Source of the dataset for non-pest plants. https://www.kaggle.com/emmarex/plantdisease 16. Tensorflow’s inception-v3 model. https://cloud.google.com/tpu/docs/inception-v3-advanced 17. Weather API. https://api.weatherbit.io/

Developing Machine Learning Based Mobile App for Agriculture …

209

18. Partitioning around medoids (PAM). https://web.archive.org/web/20111002220803/ http:// www.unesco.org:80/webworld/idams/advguide/Chapt7_1_1.html 19. Savary S, Willocquet L, Pethybridge SJ, Esker P, McRoberts N, Nelson A (2019) The global burden of pathogens and pests on major food crops. 3 20. Szegedy C, Vanhoucke V, Ioffe S, Shlens J, Wojnam S, Rethinking the inception architecture for computer vision

Attack Detection in IoT Using Machine Learning—A Survey Saeed Ali Haifa Ali and J. Vakula Rani

Abstract In the last decade, the ˙Internet Of Things(IoT) increased dramatically until it became an integral part of our daily lives. These devices are interconnected to the internet without the need for human intervention. Due to the weak configuration and unique characteristics of the internet of things has become a robust target for cyber-attack that worry the general user of these devices. Furthermore, IoT security challenges are increasing day by day and are subject to a variety of attacks. The traditional security measures, such as authentication, access control, network security, and encryption, for IoT devices and their vulnerabilities, are insufficient, ineffective, and cannot process these issues. Existing security methods must be improved to protect the IoT environment. ML/DL provided many solutions that assisted solve the challenges of the IoT and provided safety for it. The goal of this paper is to provide a study on the attacks in IoT architectures such as the sensing layer, network layer, and application layer, then present ML and DL that contributed to the solution in attack detection. In addition, we discuss the challenges of IoT architectures. Keywords Internet of Things(IoT) · Machine-Learning · Deep-Learning · Attack detection · IoT security

1 Introduction The internet of things become widely utilized nowadays, which connects devices to the internet to exchange data, IoT shares the sensor information that they gather by linking to an IoT gateway or another edge device where data is either sent to the cloud to be analyzed or analyzed locally; IoT has unique features such as resource constraints, heterogeneity, scalability, and the extensive behavior of the networks S. A. Haifa Ali (B) · J. Vakula Rani CMR. Institute of Technology, Bengaluru, Karnataka, India e-mail: [email protected] J. Vakula Rani e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_16

211

212

S. A. Haifa Ali and J. Vakula Rani

make the security and privacy in IoT more challenging and vulnerable to cyberattacks, which aims to disrupt its environment. With the advancement of technology, the interconnection of IoT ecosystems with the enterprise network and the entire Internet also occurs, making IoT and industrial IoT devices a prime target for cybercriminals and cyber-attacks. Either cause damage to devices, and files or threaten countries or stalk people either online or offline. Moreover, use them in attacks against other targets. On the other hand, there are three classes of IoT attacks that are commonly found in today’s network environment which are, access attacks that access another user account or network devices through improper means, reconnaissance attacks that are used to gather information about the target network or system, and DDoS attacks which we explain it in Table 1. However, the cyber-attacks target the IoT layers, due to the role that each layer plays, which make them subject to a diversity of attacks. The attacks can compromise the devices or other nearby devices; control the flow of the network traffic. Furthermore, attract people to trust them and then commit cyber offenses. The number of IoT devices is expected to grow by over 25.4 billion in 2030. This increase in the number of IoT devices will increase the number of attacks and the opportunities to compromise them through hacking, cracking, phreaking, denial of services attacks, online fraud, pornographic offenses, etc. Encryption and password security are one of the various vulnerabilities that are exploited by attackers. Machine Learning, is an application of artificial intelligence (AI) that provides systems the ability to automatically learn and improve from experience and Deep Learning is a type of machine learning that allows the machine to learn by itself by simulating the neurons in the human body. These algorithms can be used for detecting and mitigating the risk of attacks that come to provide solutions based on learning and prediction. ML and DL methods such as Super Victor Machine (SVM), Decision Tree (DT), RNN, DNN, Auto-encoder (Ae), etc. provide better solutions to detect and mitigate attacks. Although there are some existing surveys that have discussed attacks in IoT layers, there is a shortcoming and a lack of discussion of these types of attacks [1, 2] in depth. ˙In this study, we covered all types of promising attacks in IoT layers, and also presented some studies that addressed the promising attacks which disturb IoT security. The contribution of this paper is to concentrate on the attacks in IoT architectures, and the contribution of machine learning and deep learning to detect attacks. In addition, the challenges of IoT architecture.

2 IoT Architectures Attacks and Attacks Detection IoT architecture consists of three important layers, perception layer, network layer, and application layer, and each layer has its mission. In Fig. 1, we display the IoT architecture, which illustrates the position and the component of each layer. The perception layer is the first layer and consists of devices and controllers to sense and gather information. Then, the network layer, which is made up of connections and protocols, and the main goal is to provide a connection between networks, and finally,

Attack Detection in IoT Using Machine Learning—A Survey

213

Table 1 Type of each attack and its principle Attack

Principle

Reverse engineering attacks

A person-to-person attack in which the attacker establishes direct contact with the victim in order to persuade them to provide important information

Tampering attack

Type of physical attack in which the attacker can modify memory, obtains extra knowledge by reacting with a faulty appliance, and then try to break security

Spoofing attacks

The hackers impersonate an authorized device or user to steal data, spread malware, or bypass access control systems

Physical damage attack

Performed when the attacker is at a close distance from the device. A malicious user can take control of the computing or communication and cause damage to property and put lives at risk

RF interface attacks

Target devices that use communications methods such as Bluetooth, Bluetooth Low Energy (BLE), radio, and Wi-Fi

RFID cloning

Refer to the copying of an RFID electronic tag or smart card’s information to a cloned tag, which will then have the same features as the original tag and potentially replace it

Code and malicious

Malicious software or malware that may corrupt client computers, databases, networks, and entire clusters of servers rapidly or over time

Injection attacks

A malicious code injected into the network which retrieves all the information from the database to the attacker

Theft of data

Steal the data and then put it in computers that have been broken into, servers, or other devices with the desire to violate the privacy or get private information

Sniffing attacks

Monitor and capture all data packets passing through a given network

Browser attack

Employ a Trojan horse covertly installed on a computer system that alters the user’s web transactions in real-time, steel the messages in a public key exchange, and replaces the targeted security keys with fake ones

Denial of services (DoS)

Attacks whose goal is to flood the network or sites with unnecessary data sent by infected devices with the malicious program to control remotely, DoS causes slow services and congestion in the flow of data

Distributed senial of services (DDoS)

The attackers flood a server with internet traffic to deny users from accessing to services and websites

Social engineering

Set of tricks and techniques to deceive people and make them expose confidential and personal security information (continued)

214

S. A. Haifa Ali and J. Vakula Rani

Table 1 (continued) Attack

Principle

Sinkhole attacks

A sinkhole attack is one in which a hacked node tries to attract network traffic by announcing a bogus routing update

Sybil attack

A group of nodes impersonating different peer identities in order to compromise an IoT ecosystem and it is used to broadcast bogus data from a random network

Jamming attack

Interfere with the radio frequency of the wireless sensor node, which will jam the signal and sent the communication back to the node

Man in the middle (MiTM)

It occurs when a man or hacker intervenes in the communication between the system and the user, or between two users, in order to steal the user’s information

Phishing attack

Spoof website that pretends to be the real thing, The victims become imprisoned on the bogus website, and hackers steal all of their login information as well as credit card information

Social engineering

The attacker gains people’s trust to do things against the security or policies and fools them into providing information to get access to information

the application layer along with the user can react to the software application. The goal of this layer is to give the user permission to access and recover files. However, all three layers suffer from promising attacks and these types of attacks discuss in Table 1. The main goal of this section is to understand the risk of attacks in each layer and provide the current solutions, which have been presented by researchers with the advantage of each research. Fig. 1 IoT architecture

Attack Detection in IoT Using Machine Learning—A Survey

215

Physical Attacks: it has another two names the sensing layer and the perception layer; it is made up of sensors, and actuators. This layer in the internet of things is accountable to capture information, actuators, Zigbee, and RFID [3]. It suffers from various types of attacks that try to damage and destroyed devices such as Reverse engineering, Tampering Attacks, Jamming, Spoofing, and RF interference. Figure 2 illustrate the various type of physical attacks. The attacks enter the devices and manipulate them. Entry is through social engineering, where attacks get into devices and carry out real attacks on a large scale, such as destroying devices, eavesdropping, or other attacks. In order for physical attacks to be launched, the attacker must be close to the devices such as tampering with energy sources, endangering the communication mechanism, and reducing the life of the devices. The physical attack is nothing but a launch for the rest of the attacks like a smart lock, disabling the smart locks will potentially lead to steel money and kidnapping a member of the house. Detecting and securing this layer needs robust techniques like machine learning and deep learning methods. On the other hand, there are plenty of researchers discussing the problem of the attack on the physical layer. They have proposed many paper-based solutions to address the attacks on physical layers and one of these papers [4] has addressed the issues in the physical layer caused by jamming attacks using Random Forest (RF) machine-learning forest. The method was compared with two machine learning methods support vector machine (SVM) and K-nearest Neighbour (K-NN) in the multi-track status and one-route status. The Random Forest technique along with AdaBoost achieved significant results better than another method. Consequently, it was compared with a one RSSI status for a diverse number of nodes. Moreover, it was found that Random Forest has better accuracy in terms of testing and minimal false alarms compared to other methods. In [5] analyzed the network traffic using an Artificial Neural Network (ANN) to expose malicious intrusion in IoT devices such as smart bulbs. The result of ANN compared the performances with previous studies results that used several supervised algorithms SVM, DT, and RF. It demonstrated reasonable accuracy and error rate of ANN respectively, 77.50 and 24.48% compared to earlier results used (SVM, RF, NB, and DT). Liu et al. [6] have proposed a network invulnerable schema, called P4NIS that is supplied with three lines of defenses to detect and prevent eavesdropping attacks. Three lines of defense analyzed the packets. First encrypted the package in the upper layer with utilize cryptographic methods. Then took a program design for modifying encryption techniques, and after that, they utilized programmable forwarding policies that divided the encrypted packages into various network routes disorganized. The results showed that P4NIS rise awkwardness in eavesdropping attacks compared with modern methods. Furthermore, this schema can decrease the encryption cost by 69.85–81.24%. Shaniqua et al. [7] proposed a machine learning model using SVM to taxonomy spoofing attacks and signals received by Unmanned Aerial vehicles (UAVs), as well as, they have performed K-fold examines to improve another learning pattern via selecting various rates of K-folds then was named them K-learning methods. Moreover, the proposed work by utilizing the characteristics of GPS to be like features.

216

S. A. Haifa Ali and J. Vakula Rani

Fig. 2 Attacks in IoT architecture

Then they compared the model with five previous studies that have been carried out which scored significant accuracy, precision, recall, and F-score respectively (99, 98, 99, 98%). Pathak et al. [8] analyzed IoT security sensors to discover tampering attacks in an office environment using the decision tree (DT) method. They compiled data from the genuine world and then, use machine learning to reveal tampering attacks by using two techniques. They used a genuine-time of the traffic type to train unsupervised machine learning methods for anomaly exposure. After that according to traffic, labels were created, and the decision tree supervised techniques has been used. Moreover, the result based on the Decision tree scored high accuracy of 91.61% and a low false-positive rate (FPR). Network Attacks: This layer is used to convey data from the lower layer to the upper layer for processing [9, 10]. The major threats here are eavesdropping attacks, a man in the middle (MITM), Sybil attacks, Sinkhole attacks, DDoS attacks [11], and RFID Cloning. When the attackers hack this layer, they turn IoT devices into botnets. The aim of Network Layer attacks is to interrupt the route between the source and destination. This is one of the routing protocols that has been chosen. In network attacks, the attackers bombard the network with more traffic through

Attack Detection in IoT Using Machine Learning—A Survey

217

compromised nodes or fake nodes, and from here, they allow an attack called Sybil attacks. Tampering with devices might allow attacks to make changes in the routing table and security key that would affect upper layers. On the other hand, the network layer is the most important layer due to its position in the middle between the physical layer and the application layer, therefore, securing this layer is essential and many proposed works addressed the issues in this layer. We present some proposed solutions based on machine learning and deep learning, which have been provided to solve the challenges in the network layer. Kiran et al. [12] have built a model to identify Man-in- the- Middle attacks in the IoT network layer using machine-learning methods such as Naïve Bayes (NB), Support Vector Machine (SVM), and Adaboost. They built a testbed to simulate an IoT environment using a node, sensor, and wireless router and they built an adversarial system using a laptop. On the other hand, they collected data from sensors and transmitted it to ThinkSpeak platform using the wireless gateway. After that, the sensor value has been captured by a node and then transmitted to ThinkSpeak server, which was stored. At the last, the performance measurements of the classifier scored high of 100% in terms of accuracy, detection rate, and false alarm rate (FAR). Newaz et al. [13] have proposed an Intrusion Detection System scheme, to demonstrate different vulnerabilities of personal medical device communication and then implemented five diverse cyber-threats such as man in the middle, Replay, false data injection, and DoS on commercially available healthcare devices using machine learning SVM, KNN, DT, RF. Then, they presented a method called HEKA, to watch and reveal cyber-threats in the network traffic. They have implemented the method in a testbed that contained eight medical appliances and estimated its accomplishment with four various attacks. Furthermore, the proposed showed that HEKA is successful in revealing diverse attacks with a score of e 98% of accuracy and a F1 score of 98%. Farzaneh et al. [14] have presented a deep learning method called the Fuzzybased method that used three metrics to detect Repair Attacks on routing protocol. Moreover, the result used the Cooja simulator in the operating system Contiki and showed that the method was able to detect the Repair attack and scored a high True Positive Rate (TPR) and minimal False Positive Rate (FPR). Alaiz-Moreton et al. [15] have proposed a model to reveal and detect attacks like DOS, MiTM, and Intrusions in IoT environments by using an intrusion detection system (IDS). The work concentrated on making classification models that can feed an IDS using a dataset containing frames under attacks of an IoT system that utilized the MQTT protocol. They addressed two kinds of methods for taxonomizing the attacks, ensemble methods, and deep learning models. SVM, RF, Boosting Gradient, XGBoost, and LSTM have been used in this model and achieved a high score accuracy of over 90%. However, SVM achieved a result better than the other methods. Anthi et al. [16] have proposed a three-layer Intrusion Detection System (IDS), which used supervised machine learning methods to detect network cyber-attacks (Denial of Service (DoS), Man-In-The-Middle (MITM), Spoofing, and Replay attack) on IoT networks. The system was assessed by training eight devices on a testbed. The system included three main functions, which are, to taxonomy the type

218

S. A. Haifa Ali and J. Vakula Rani

as well as, profiled the natural action of IoT devices that are linked to the network, then detected wireless attacks against connected IoT devices, and after that, classified the attacks that have been deployed. The proposed achieved a high score in the achievement of the three tasks resulting in an F-measure of: 96.2, 90.0, and 98.0%. Application Attacks: The application layer manages connecting end-users to endpoints using a user interface and deals with a large number of data transactions [17, 18]. The most challenging layer to be protected is the application layer. Plenty of the vulnerabilities found here are based on sophisticated user inputs that are complex to express with intrusion detection. Moreover, is vulnerable to software attacks such as malware, viruses, worms, Sniffing attacks, Phishing attacks, Injection attacks, and Browser attacks. This layer is also accessible and exposed to the world. SQL injections, a type of application attack, were responsible for breaches of data in the year 2014. It is the third most prevalent attack, coming after malware and DDoS. Other common vulnerabilities in this layer are included, such as security misconfiguration. Attackers have the ability to change application information and capture private data without being exposed by network protection mechanisms. The researchers presented many types of researches based on application attacks and we will present a few of them. Azmoodeh et al. [19] have proposed a neural network (NN) deep learning technique to detect Internet of Things malware in the IoT device’s Operational Code using the deep learning method. They transferred OpCodes in a vector space and applied deep Eigenspace learning methods to classify malicious and benign applications. The proposed achieved an accuracy of 98.37% and a precision rate of 98.59%. Mao et al. [20] have presented a learning-based aggregation analysis mechanism to detect phishing pages and determined the similarity of page layouts using machinelearning methods SVM, DT, and RF. In addition, the result of the classifier achieved above 95% accuracy. Then they compared the proposed work with other approaches and found that their work achieved better performance. Moti et al. [21] have presented a structure to detect malicious patterns called MalGan by using the deep learning method Generative Adversarial Network (GAN), the performance analysis was carried out using VX-Heaven dataset, which contained malicious and normal software. The proposed MalGan scored a higher detection rate compared to four previous works of malware detection algorithms in terms of accuracy, F1, testing time, and training time. Shafiq et al. [22] have proposed a feature selection approach named CorrAUC to detect Malicious Bot-net attacks, then relied on CorrAUC, they developed a feature selection method called Corrauc using machine learning RF, NB, SVM, and C4.5 DT. They applied two methods called TOPSIS and Shannon Entropy based on a bijective soft set to advocate selected features for malicious traffic in the IoT network. They used the Bot-IoT dataset to carry out their performance. The proposed showed a significant result of over 95%. Many feature selection approaches have been presented in [23], which increases accuracy, noise reduced. Sarkar et al. [24] presented algorithms to determine the key parameters necessary to follow and sniff a Bluetooth Low Energy (BLE) connection in the connected

Attack Detection in IoT Using Machine Learning—A Survey

219

state by concentrating on the adversary setting with a low-cost single radio and developing a suite of real-time algorithms. The algorithm was implemented in the open-source platform. At last, the performance of the implemented method achieved better accuracy and over 80% as well as, stability to BLE operational dynamics. The types of attacks in each layer are presented in Fig. 2, in which we demonstrate twenty attacks that break into the devices either in the perception layer or, hacking and, phreaking the network layer or grabbing data and destroying files in the upper layer either closely or remotely. In Table 1, we display the diverse types of attacks in IoT architecture, and these attacks threaten the general security of each layer and are considered one of the most important security concerns that harm and impede the work of each layer, The table shows the principle and the goal of each attack in each layer to make it effortless concept for the other researchers. We observe from the table that all the types of attacks aforementioned concentrated on hacking either by breaking into devices or communication to damage and sabotage. Moreover, these attacks focus on fooling people like the goal of social engineering either online or offline. on the other hand, we notice that the most common hackers target E-mails is sniffing and phishing attacks by sending lots of tricking emails such as advertised emails or impersonating to obtain private details about the address, name, and bank account detail or ID. In Table 2 we summarize the three layers (perception, network, and application) in terms of the types of attacks in layers that disturb the security, we enhance that with the main purpose and primary task of each layer, furthermore, we present the weaknesses and the challenges of each layer that make them vulnerable to attacks. Through our research on weaknesses in IoT layers security, we observe that not securing devices in the lower layer, the reason behind that is the natural characteristics in IoT make devices vulnerable to hackers and the amount of data flowing upper layer, which make the network overcrowding. The Physical layer-based security solutions are listed in Table 3, which consists of a few studies references regarding the attacks in the perception layer, we display the contribution of each study and which type of attack the researcher concentrated on and analyse, and we present the techniques that the researchers use it to solve the problem, we pay attention that the majority of researchers are interested to solve spoofing, jamming and malicious attacks in perception layer. In fact, these types of attacks are the most dangerous in the lower layer that disturbs the devices. Finally, we extracted the advantages of each study that most of them got high accuracy and detection rate. The network layer-based security solutions are listed in Table 4. The table contains some studies regarding the attacks in the network layer, we discuss the contribution of each study and the attacks that the researchers concentrated on, and we display the techniques that the researchers use to solve the problems, we observe that the majority of researchers are focusing on DoS, DDoS, MiTM attacks and intrusion. Indeed, these types of attacks are critical that affect the task of this layer; we extract the advantages of each study that most of them got high accuracy and precision, and less false positive rate.

220

S. A. Haifa Ali and J. Vakula Rani

Table 2 Illustrates the attacks, the main purpose of each layer, and the challenges Layer

Attacks

Main purpose

Challenges

Physical layer

• • • • • • • • •

Reverse engineering Tampering Social engineering Jamming Spoofing DOS Fake node Physical damage RF interference

Gather data

• IoT is unmonitored and vulnerable to hackers • Destroying perception devices and manipulating gathered information • Resource constrains devices • Data privacy

Network layer

• • • • • • • •

Man in the middle DDOS Data transfer attack RFID cloning Spoofing Sinkhole attack Sybil attack Routing information attacks

Transmit collected data

• TCP and IP protocol • Concentrate on the influence of network resource availability • Wireless networks due to devices connected to IoT networks via wireless communication • Due to a high volume of data, there is network congestion

Application layer

• • • • • •

Code and malicious Injection attacks Theft of data Sniffing attacks Phishing attacks Browser attack

User-requested assistance

• Because it deals with a high volume of data transfers, it presents a huge issue for IoT security • Attacks against software (data privacy, control of access, and leakage of data)

Table 5 shows Application layer-based security solutions. The table is composed of several studies related to the attacks in the upper layer, we overview the contribution of each study and the attacks that the researchers work on solving, and we show the techniques that contribute to solving the problems, we observe that the researchers are focusing on malware and phishing attacks. These types of attacks are significant due to the task of the layer such as dealing with data and files; we summarize the advantages of each study that most of them got high accuracy and precision, high recall, and f1.

3 Challenges Although we discussed the solutions in IoT regarding machine learning and deep learning, the potential challenges [37], have not been solved by these methods. One of these problems is securing the use of data, leaving the devices without shutting

Year

2019

2019

2019

2017

2017

Author and References

Upadhyaya et al. [4]

Khatun et al. [5]

Goel et al. [25]

Aref et al. [26]

Han et al. [27]

Jamming

Attack

Proposed a PHY-layer authentication to minimize the connection between security factors and landmarks and then supply security based on the scheme

Proposed a method called multi-agent reinforcement learning algorithm, based on Q-learning, to bypass jamming threats

Presented the cooperation between DNN methods and block-chain security that can create a model to eliminate tampers

Spoofing

Jamming

Adversarial tampering

Presented Artificial Neural Malicious nodes Network (ANN), which can be carried out to process the traffic in the network to be able to detect malware in IoT

Discussed machine-learning methods to detect jamming detection using data from emulation and genuine networks

Contribution

Table 3 Physical layer attacks solutions

ML

Q-learning

DNN

ANN

RF, real AdaBoost

Countermeasure method

FAR, miss DR

Timing

High accuracy

(continued)

Reasonable accuracy and error rate

High accuracy

Advantage

Attack Detection in IoT Using Machine Learning—A Survey 221

Year

2017

2017

Author and References

Xiao et al. [28]

Shi et al. [29]

Table 3 (continued) Jamming

Attack

Proposed a device with free Human daily activities user authentication by essencing physiological and features established in everyday activities

Created a jamming detection connection game that takes advantage of user mobility to mend the signals against jammers

Contribution

DL

Deep reinforcement learning

Countermeasure method

High accuracy

Fast convergence rate

Advantage

222 S. A. Haifa Ali and J. Vakula Rani

Year

2020

2020

2020

2017

2019

2019

Author and References

Kiran et al. [12]

Newaz et al. [13]

Zaman et al. [1]

Singh et al. [30]

Roopak et al. [31]

Newaz et al. [13]

Attack

MITM, replay attack, false data injection, DoS

DDOS

Dos, MITM, Intrusions

Discussed the three-layer with DOS, MiTM, spoofing, the use of supervised methods reconnaissance, and replay to detect cyber-attacks on IoT network

Suggested four deep learning models with the utilized of CICIDS2017 datasets and compared them with machine learning algorithms

Concentrated on creating classification models that can feed an IDS using a dataset containing frames under attacks of an IoT system that uses the MQTT protocol

Proposed a Fuzzy method to Local repair attack detect and reveal of Repair Attack on the routing protocol

Proposed an Intrusion Detection System (IDS), which is called HEKA, to watch medical appliances traffic and detect attacks

Have built a model to identify Man-in-the middle Man-in- the- Middle attacks in IoT network layer

Contribution

Table 4 Network layer attacks solutions

ML

DL

SVM, RF, Boosting Gradient, XGBoost, LSTM

DL, Fuzzy Logic

SVM, KNN, DT, RF

SVM, NB, DT, Adaboost

Countermeasure method

(continued)

High F1 measures

High accuracy, high precision

High accuracy

High TPR, less FPR

High accuracy, F1

Misclassification, High accuracy

Advantage

Attack Detection in IoT Using Machine Learning—A Survey 223

Year

2017

2017

Author and References

Mohamad et al. [32]

Li et al. [33]

Table 4 (continued) Attack

Countermeasure method

Proposed a Sybil attacks detection in WSN by taking advantage of the connection between path responses from various sensors

Sybil attack

DL, fuzzy Logic

Suggested IDS named Routing attacks namely, hello ML compression header processor flood, sinkhole, and wormhole that examines 6LoWPAN to attacks minimize the routing attacks

Contribution

High accuracy

High accuracy

Advantage

224 S. A. Haifa Ali and J. Vakula Rani

2019 Conveyed the Opcodes into a vector space and apply fuzzy then rapid fuzzy tree methods to detect malware and its type

2020 Presented an IoT Phishing attacks, DDoS attacks CNN, LSTM micro-security add-on hosted in the device using a CNN model to detect URL-based attacks directed to a client’s IoT devices. and LSTM model hosted at the back-end servers for detecting botnet attacks in IoT devices

2021 Proposed an anomaly detection DOS, DDOS, Reconnaissance, DNN technique utilizing the deep and information theft attacks neural network for the IoT network layers to classify traffic as benign and unbenign

Parra et al. [35]

Ahmad et al. [36]

Malware

SVM, DT, KNN, RF

SVM, DT, AB, RF

Dovom et al. [34]

Phishing

2019 Aimed to enable automated page-layout-based phishing detection techniques using machine learning techniques

DL

High accuracy

High accuracy

High accuracy, high precision, high recall

High accuracy, reasonable F1

High accuracy, high precision

Countermeasure method Advantage

Mao et al. [20]

Malware

Attack

2018 Presented an IoT botnet malware detection method based on the selection of Opcodes as a feature for classification mission

Contribution

Azmoodeh et al. [19]

Author and References Year

Table 5 Application layer attacks solutions

Attack Detection in IoT Using Machine Learning—A Survey 225

226

S. A. Haifa Ali and J. Vakula Rani

them out could be subject to hackers that can exploit the power and get private information, for this, it should secure network connection to prevent any intrusion to gain access to devices. On the other hand, with data leaks in the IoT ecosystem, hackers can access the data analyzed by the system. This may have private data like location, bank account data, and health registry. Nevertheless, taking advantage of insecure connections is not the only thing for hackers to get beneficial data. The entire data is sent and stored in the cloud, and then cloud-hosted services are in danger of exterior attacks. Therefore, breaches of data are possible from the devices themselves and the cloud environments to which they are connected. For machine learning and deep learning, which are considered as a solution to IoT security, they suffer from selecting the proper ML and DL algorithm for the mission [38]. Due to IoT being considered resource constraints, this problem limits the use of DL and ML algorithms to protect IoT networks. Finally, DL needs a huge number of training data to do the task.

4 Conclusion IoT becomes crucial in our life nowadays; however, it is subject to attacks that disturb security. Securing IoT is difficult and traditional security cannot process IoT but can handle other networks. For this, machine learning and deep learning can detect and mitigate the risk of attacks. In this survey, various attacks type on the tree layer, physical, network and application layer have been discussed. We covered security attacks on the three layers using machine learning and deep learning methods. This survey targets to provide a study of the current solution of machine learning and deep learning that has been enhanced with previous researches.

References 1. Zaman S et al (2021) Security threats and artificial intelligence based countermeasures for internet of things networks: a comprehensive survey. IEEE Access 2. Lin J et al (2017) A survey on internet of things: architecture, enabling technologies, security and privacy, and applications. IEEE ˙Internet of Things J 4(5):1125–1142 3. Liang X, Kim Y (2021) A survey on security attacks and solutions in the IoT network. In: 2021 IEEE 11th annual computing and communication workshop and conference (CCWC). IEEE 4. Upadhyaya B, Sun S, Sikdar B (2019) Machine learning-based jamming detection in wireless iot networks. In: 2019 IEEE VTS Asia Pacific wireless communications symposium (APWCS). IEEE 5. Khatun MA, Chowdhury N, Uddin MN (2019) Malicious nodes detection based on artificial neural network in IoT environments. In: 2019 22nd International conference on computer and ınformation technology (ICCIT). IEEE 6. Liu G et al (2021) Softwarized iot network immunity against eavesdropping with programmable data planes. IEEE Internet of Things J 8(8):6578–6590 7. Shaniqua A, Mehmood A, Elhadef M (2021) Detecting signal spoofing attack in uavs using machine learning models. IEEE Access 9:93803–93815

Attack Detection in IoT Using Machine Learning—A Survey

227

8. Pathak AK et al (2021) Anomaly detection using machine learning to discover sensor tampering in IoT systems. In: ICC 2021-IEEE ınternational conference on communications. IEEE 9. Mrabet H et al (2020) A survey of IoT security based on a layered architecture of sensing and data analysis. Sensors 20(13):3625. 10. Hassija V et al (2019) A survey on IoT security: application areas, security threats, and solution architectures. IEEE Access 7:82721–82743 11. Amrish R, Bavapriyan K, Gopinaath V, Jawahar A, Vinoth Kumar C (2022) DDoS detection using machine learning techniques. J IoT Soc Mobile Analytics Cloud 4(1):24–32 12. Kiran KVVNLS et al (2020) Building a intrusion detection system for iot environment using machine learning techniques. Procedia Comput Sci 171:2372–2379 13. Newaz AKMI et al (2020) Heka: a novel intrusion detection system for attacks to personal medical devices. In: 2020 IEEE conference on communications and network security (CNS). IEEE 14. Farzaneh B et al (2020) A new method for intrusion detection on RPL routing protocol using fuzzy logic. In: 2020 6th International conference on web research (ICWR). IEEE 15. Alaiz-Moreton H et al (2019) Multiclass classification procedure for detecting attacks on MQTT-IoT protocol. Complexity 2019 16. Anthi E et al (2019) A supervised intrusion detection system for smart home IoT devices. IEEE Internet of Things J 6(5):9042–9053 17. Tawalbeh L et al (2020) IoT privacy and security: challenges and solutions. Appl Sci 10(12):4102 18. Asharf J et al (2020) A review of intrusion detection systems using machine and deep learning in internet of things: challenges, solutions and future directions. Electronics 9(7):1177 19. Azmoodeh A, Dehghantanha A, Choo KKR (2018) Robust malware detection for internet of (battlefield) things devices using deep eigenspace learning. IEEE Trans Sustain Comput 4(1):88–95 20. Mao J et al (2019) Phishing page detection via learning classifiers from page layout feature. EURASIP J Wirel Commun Netw 2019(1):1–14. 21. Moti Z et al (2021) Generative adversarial network to detect unseen internet of things malware. Ad Hoc Netw 122(2021):102591 22. Shafiq M et al (2020) CorrAUC: a malicious bot-IoT traffic detection method in IoT network using machine-learning techniques. IEEE Internet of Things J 8(5):3242–3254 23. Abdulwahab HM, Ajitha S, Saif MAN (2022) Feature selection techniques in the context of big data: taxonomy and analysis. Appl Intell 1–46 24. Sarkar S, Liu J, Jovanov E (2019) A robust algorithm for sniffing ble long-lived connections in real-time. In: 2019 IEEE global communications conference (GLOBECOM). IEEE 25. Goel A et al (2019) DeepRing: protecting deep neural network with blockchain. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops 26. Aref MA, Jayaweera SK, Machuzak S (2017) Multi-agent reinforcement learning based cognitive anti-jamming. In: 2017 IEEE wireless communications and networking conference (WCNC). IEEE 27. Han G, Xiao L, Vincent Poor H (2017) Two-dimensional anti-jamming communication based on deep reinforcement learning. In: 2017 IEEE international conference on acoustics, speech and signal processing (ICASSP). IEEE 28. Xiao L, Wan X, Han Z (2017) PHY-layer authentication with multiple landmarks with reduced overhead. IEEE Trans Wireless Commun 17(3):1676–1687 29. Shi C et al (2017) Smart user authentication through actuation of daily activities leveraging WiFi-enabled IoT. In: Proceedings of the 18th ACM ınternational symposium on mobile Ad Hoc networking and computing 30. Singh R, Singh J, Singh R (2017) Fuzzy based advanced hybrid intrusion detection system to detect malicious nodes in wireless sensor networks. Wirel Commun Mobile Comput 2017 31. Roopak M, Tian GY, Chambers J (2019) Deep learning models for cyber security in IoT networks. In: 2019 IEEE 9th annual computing and communication workshop and conference (CCWC). IEEE

228

S. A. Haifa Ali and J. Vakula Rani

32. Napiah MN et al (2018) Compression header analyzer intrusion detection system (CHA-IDS) for 6LoWPAN communication protocol. IEEE Access 6:16623–16638 33. Li Q et al (2017) Channel-based sybil detection in industrial wireless sensor networks: a multi-kernel approach. In; GLOBECOM 2017–2017 IEEE global communications conference. IEEE 34. Dovom EM et al (2019) Fuzzy pattern tree for edge malware detection and categorization in IoT. J Syst Architect 97:1–7 35. Parra GDLT et al (2020) Detecting Internet of Things attacks using distributed deep learning. J Netw Comput Appl 163:102662 36. Ahmad Z et al (2021) Anomaly detection using deep neural network for IoT architecture. Appl Sci 11(15):7050 37. Kamel DK (2021) Wireless IoT with blockchain-enabled technology amidst attacks. IRO J Sustain Wirel Syst 2(3):133–137 38. Amrish R, Bavapriyan K, Gopinaath V, Jawahar A, Vinoth Kumar C (2022) DDoS detection using machine learning techniques. J IoT Soc Mob Analytics Cloud 4(1):24–32

An Extensive Study on Logic Emerging IoT Adiabatic Techniques for Low-Power Circuit T. Vijayalakshmi and J. Selvakumar

Abstract A low-power, energy-efficient circuit is essential for IoT edge devices, which increasingly perform data-intensive applications. Nanometer technology nodes push standard CMOS to its limits, which include increased leakage and increased power consumption. Appropriate algorithms for low-power circuits include adiabatic logic and approximation computing. It is possible to construct circuits that are more energy efficient by using adiabatic logic. The adiabatic logic’s dual-rail construction and power clock approach, on the other hand, increase the overall footprint. More power is conserved by lowering the circuit’s complexity and size while utilising approximation computing. For the Internet of Things (IoT), energy efficiency, and security, adiabatic circuits have the potential to work together. IoTRF-powered devices can benefit greatly from adiabatic circuits even though they have been around for more than six decades, as demonstrated by some of the recent advancements. These enhancements are described in detail, with an emphasis on the main design challenges and opportunities associated with adiabatic circuits. Keywords Adiabatic circuits · Nanometer · Internet of Things · Power clock · Energy consumption

1 Introduction IoT refers to a network of interconnected devices that communicate data in order to facilitate the development of intelligent software. Smart homes, smart cities, and health care all use IoT strategies to monitor, anticipate, and regulate the world around us, which is a huge benefit. Mobile phones and wearables such as thermometers and moisture sensors (to name a few) are just a few examples of the wide variety of devices T. Vijayalakshmi (B) · J. Selvakumar Department of ECE, SRMIST, Kattankulathur, Chennai, India e-mail: [email protected] J. Selvakumar e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_17

229

230

T. Vijayalakshmi and J. Selvakumar

that fall into this category [1–3]. Data is collected by IoT devices and transferred to the cloud for processing. As a result, the network infrastructure is under greater strain as a result of the increasing volume of IoT edge devices and the data they generate. By 2022, there will be more than 28 billion IoT devices, and by 2025, there will be 75.44 billion [4, 5]. As a result, a novel paradigm for dealing with the massive amounts of data produced by IoT edge strategies is required. An emerging computing example known as edge computing is one in which computation takes place at the network’s edge. In the case of edge, nodes are located close to the end-user. As a result, they are less influenced have a low latency. As a result, the cloud server receives less sensitive data [5]. Consequently, researchers can use edge computing to create [6]. In order to create deep learning at the edge, IoT devices will need to have more processing power, which will increase power dissipation. In addition, the computing and energy resources of most IoT devices are constrained. As a result, IoT devices. Figure 1 shows the advantages of using adiabatic logic and approximation computing to create low-power and energy-efficient Internet of Things devices. Since battery life is a concern for mobile devices, the adiabatic approach is an absolute necessity, where adiabatic logic can be used to solve the need for long battery life in portable electronics [7]. It has been discovered that the most promising approach to increasing the battery life of integrated circuit devices is adiabatic logic. A long-lasting battery life is needed because of the rise in demand for handheld devices. More switching activity can be expected due to the fact that today’s ICs operate at extremely fast speeds. The significant power dissipation means that big heat sinks are required. As a result, the gadget must be larger. Because of this, we must devise methods that utilise the least amount of energy possible. When it comes to portability, the biggest obstacle is power consumption. In a conventional CMOS circuit, power dissipation can take the following forms: (1) Dynamic Power Dissipation, which refers to the power expended by a device while it is switching. Static Power Dissipation (SPD) is the unpowered [8]. In the VLSI design process, there are a variety of ways to reduce the amount of power consumed. The scale of the supply voltage has also been altered to reduce power consumption. As a result, the circuit’s speed is reduced when the supply voltage is reduced.

2 Background on Adiabatic Circuits Irreversibility and heat generation were ideas that physicist Landauer considered in the 1960s when introducing adiabatic circuits for computers [9]. Adiabatic operations are defined as those that do not raise the environment’s entropy during their execution [10]. Adiabatic operation can be difficult and impossible to achieve because of the sluggish current flow and inherent losses in the process in standard CMOS devices [11–13]. However, new logic families have been proposed to take advantage of adiabatic operation to reduce power consumption.

An Extensive Study on Logic Emerging IoT Adiabatic Techniques …

231

Fig. 1 Uniting the estimate computing and logic

A trapezoidal or sinusoidal power supply signal is the basic characteristic of these logic families, even if they range greatly in their ability to achieve full adiabatic operation. Since it synchronises the flow of data and is sometimes mentioned to as the power-clock signal, this signal also serves as the adiabatic circuit’s clock [14]. Figure 2 depicts the adiabatic equivalent circuit for a given operation. The transistor’s on-resistance and interconnect resistance are represented by R and C, respectively, while the output wire’s capacitance is represented by C. A time of tr is the power supply signal. Power loss across R can be reduced if time r is long enough to keep up with the time constant of the resistors in the circuit. When charging and discharging are taken into account, the total switching energy dissipated each cycle is calculated. swi =2 E ad

RC 2 C Vdd tr

(1)

Unlike normal static CMOS based process, in adiabatic process, a longer transition period lessens the overall switching energy, as indicated by (1). at which point in time the  switching energy  required by static CMOS process has reached a critical 2 value E st = 21 αC Vdd is Energy consumed by adiabatic operation is equal to E st and can be found by comparing it to (Eq. 1). trcrit = 4

RC α

(2)

where is a measure of the amount of switching activity. As a result, adiabatic circuits use less switching energy than traditional circuits when tr is greater than trcrit . As a result, adiabatic operation is well-suited to applications with low operating moderate

232

T. Vijayalakshmi and J. Selvakumar

Fig. 2 Supply signal versus traditional charging with a continuous DC voltage

to high activity levels. At low frequencies, leakage energy can take upsurge overall energy consumption in adiabatic operation, thus it’s crucial to keep this in mind. Because of this, a particular frequency has a significant impact on the amount of energy used overall [15, 16]. It can range from a few thousandths of a kilohertz to tens of thousands of kilohertz, depending on the technology. Because the output charges and recuperates each time the power supply signal is applied, it is practicable to power gate adiabatic logic circuits during idle mode to reduce the amount of energy that is dissipated. The power-clock generator’s design characteristics should be considered when developing the power-gating circuitry.

3 Recent Developments 3.1 Adiabatic Circuits for RF-Powered IoT Applications RF power harvesting IoT devices could benefit greatly from adiabatic circuits. RFIDbased systems, which have traditionally had limited computer capabilities, are two examples of these applications. It is possible to adiabatically drive a digital logic in these devices that is already powered by RF because the wirelessly gathered signal is already in sinusoidal form For one thing, the difficulties of generating a power clock signal are lessened, significant power loss associated with the rectification process is eliminated, due to adiabatic operation when using this approach to improve RF-powered logic’s energy efficiency, which is a key concern in this design. Wireless link based on near-field inductive link for adiabatically powering an 8-bit ALU built in several families of adiabatic logic utilising 65 nm technology [17]. At 13.56 MHz, the wireless link transmits power at a dBm level of 24. At a distance of 6.5 cm, the power efficiency drops to − 37.4 dB, the simulated findings show a reduction in power usage of up to 30. Pass transistor adiabatic logic (PAL) required two out-of-phase power-clock signals to achieve this substantial decrease in power

An Extensive Study on Logic Emerging IoT Adiabatic Techniques …

233

Fig. 3 In the case of a wirelessly powered application, directly harvesting from the wireless link the requisite power-clock signals for adiabatic logic

[17]. According to Fig. 3, two receiving coils with a 180-degree phase difference capture these two signals directly. In order to assure proper operation, the negative components of the wirelessly acquired signal must be eliminated. Figure 3 shows an example of a capacitor signal shaper that can accomplish this goal [18]. The output nodes of PAL stay fluctuating for a brief amount of time during process, which reduces the robustness of the architecture. As ECRL is more robust and allows for lower voltages, we’ve looked into using it for RF-powered applications as well [19]. ECRL, on the other hand, necessitates a 90° phase difference between each of the four power-clock signals. With passive LC components, a phase shifter is needed [20]. In order to reduce resistive loss, the size of these passive devices is increased, especially at low frequencies. For wireless power harvesting, existing adiabatic logic families have intriguing tradeoffs.

3.2 Adiabatic Logic Enhanced Hardware Security Adiabatic circuits are more resistant to side- attacks in terms of hardware security. The lower SNR of these devices reduces the quantity of side-channel leakage [21, 22]. During the adiabatic operation recovery phase, there is a way for leaking. pMOS transistor’s turn off when the output voltage approaches threshold. Consequently, during the subsequent charge cycle, less current from the power source flows, allowing information on the prior input signals to leak out. The adiabatic operation is dependent on the use of ECRL with four-phase power clock signals developed in 65 nm technology. To put it another way: When using, the encryption efficiency (measured in Kb/sec/W) is boosted by roughly five. By reducing throughput by 18% and increasing physical area slightly, this large improvement in efficiency was made possible (2%). In addition, we used a correlation power analysis

234

T. Vijayalakshmi and J. Selvakumar

(CPA) approach to figure out an adiabatic SIMON core’s power-based side channel resistance [23–25]. As a result, we calculated correlation coefficients. Static CMOSbased unprotected SIMON core has an MTD of 1354 while unprotected adiabatic version has an MTD of 5718. It is therefore more than four times more resistant to power-based side-channel examination assaults if the adiabatic implementation is used in place.

4 Avoid Common Mistakes to Adiabatic Logic Design 4.1 Don’t Use Diodes True adiabatic design is all about avoiding diodes at all costs. It’s essential to remember that if a diode is used as an essential functional element in a circuit, it will have to be substituted in the future as energy efficiency requirements grow more stringent. This is especially true when compared to, for instance, junction diodes, because they are inherently thermodynamically irreversible and generate an irreducible amount of entropy whenever they are used in that capacity. There is a built-in voltage drop across diodes, and this “diode drop” leads in an irreversible loss of QV for a given charge Q carried across the diode. For example, In other words, a dissipation-less diode is equivalent to an “Energy Maxwell’s demon Hamiltonian dynamics laws embedded in all modern physics through quantum mechanics [26]. Diodes were commonly utilised in the charge return path of early adiabatic circuit designs, starting with Watkins. Until we can achieve even greater energy efficiency, this strategy may be beneficial if the diode drop is smaller than logic voltage swings.

4.2 Rules of Don’t Disobey Transistor Even if diodes are non-adiabatic, transistors can still be used for adiabatic operation as long as two basic principles are adhered to, even though they’re not ideal switches: • Never switch on a transistor if its source and drain terminals have a considerable voltage differential. • When a transistor’s channel is carrying a considerable amount of current, it should never be turned off. As for the first rule, it should be self-evident. For example, we know that a dissipation of 12CV2 occurs when a dynamic node of capacitance C is connected directly to a static reference signal of voltage that differs by V. This dissipation is still 14CV2 when they converge to their average level even when both nodes are isolated and have capacitance C. Turning on the transistor results in a continuous power dissipation if the two nodes are coupled to voltage sources that are different from each other. It is

An Extensive Study on Logic Emerging IoT Adiabatic Techniques …

235

generally impossible to verify that the voltages before the transistor is turned on are exactly the same in most adiabatic logic types because of noise issues, leakage, etc. However, we should strive to come as near to a match as feasible. In fact, when the gate-to-source voltage passes a certain threshold, transistors aren’t ideal switches, and they don’t immediately transition from “perfect-on” to “perfect-off” (however slowly). A lossless diode can be built using perfect switches that are thermodynamically impossible [26]. In SCRL, however, the circuit layout may be easily modified so that this behaviour is avoided [27] by connecting the gate of a third complementary-type transistor to the gate of the outer FET in parallel with the inner FET. In order to minimise unwanted non-adiabatic dissipation while charging the source node, this second FET’s strong on-state diverts the source node’s current from the FET that is turning off, resulting in a slight voltage drop across it. Dual rail logic, which enables the designer to provide full-swing charging paths everywhere, can be used to avoid the rule being violated in inverted Boolean gate pull-up/pull-down networks of any desired complexity. However, this scenario demonstrates that even if one is aware of the rule (2), one might readily break it by unintentionally. If gate-activated or source-activated transistor switch offs are ever used, it is imperative that designers be made aware of and emphasise the adiabatic circuits rule. Voltages at both the source and drain nodes must remain constant or fluctuate along identical paths when the transistor is turned off.

4.3 Use Reversible Logic A large number of so-called adiabatic logic styles are not expressly nor implicitly irreversible. A direct proof of the link can be found in modern physics’ most fundamental and well-known rules. The Schrödinger wave equation is a fundamental part of the model’s foundation as a quantum theory. To demonstrate the reversibility (bi-objectivity) of the time-evolution, any Hamiltonian dynamical system’s state variables (i.e., the amplitudes of wave functions in quantum theory) are subject to a first-order differential equation. Even seemingly irreversible occurrences such as are explained away in pure quantum theory as predictable emergent phenomena predicted by a totally reversible underlying theory [28]. No deviations from this micro-reversibility have ever been seen. Reversible evolution of a microstate in interaction with an unknown environment cannot be tracked precisely by the modeller, and this results in macroscale irreversibility. The link between adiabatic and reversible logic is obvious when considering the reversibility of mainstream quantum physics. So, for example, let’s say that an algorithm claims it can operate on a bit that can take one of two possible values and convert that bit to one with a single, unambiguous value. The number of possible physical states will not be reduced by this action. Because the logical content of the bit no longer distinguishes the states, other physical factors, such as the thermal vibrational state of atoms in the immediate environment,

236

T. Vijayalakshmi and J. Selvakumar

must now be used to separate them. Physical information in the system as a whole has not been altered at all. After erasing a given bit’s state, the information about the given bit’s state that was previously known is now unknown, which is exactly what entropy is: unknown information. So the total entropy has risen, as a result of this. State space size has risen by 2 times, hence this is a logarithm of the increase in state space. This value of entropy is equivalent to the Boltzmann constant k ln 2 when expressed in logarithm base e units. Semi-adiabatic techniques, on the other hand, do not scale to ever-lower levels of dissipation when technical improvements or speed decreases, even if they are not conceptually reversible.

4.4 Don’t Over-Constrain the Design A number of logic systems have been proposed for creating large-scale circuits that approach true adiabaticity (and consequently total logical reversibility). The costs (in terms of space, time, and energy) of doing certain computations with existing schemes are asymptotically higher by unboundedly huge factors than with alternative schemes. Adiabatic logic families [29], don’t allow reversibility over numerous levels of sequential, pipelined logic. The SCRL project of Younis and Knight [30, 31] aimed to address this specific issue. A pipelined, reversible, fully adiabatic logic was provided by them. SCRL, on the other hand, turns out to be overly restrictive because it necessitates a transition for every logic node on every clock cycle. When implemented in SCRL, even memory that are simply storing data conduct active logic transitions for each stored bit [32]. Since SCRL does not enable stored bits to remain idle, it is asymptotically less cost-effective than an alternate adiabatic technique.

5 Conclusion For long-term decades of cost-effective computing, which provides necessitates for a close approach to the real physical limits of computing. This is especially true when considering the requirement for nearly total physical (and thus logical) reversibility of the computing mechanism. These constraints rule logic circuit designs and transistors (or other current switches) must be monitored for their current carrying status when the device is turned off, as well as their voltage state when the device is turned on. Another requirement is that hardware algorithms of the highest asymptotically efficient performance can be expressed using this logic family. In the coming century, this forecast is as certain as the confluence of modern physics’ and technology’s juggernaut march. Meeting all of these conditions will be an absolute economic necessity. None of the adiabatic logic families created by other researchers have met all of these criteria. An order of magnitude improvement in energy efficiency

An Extensive Study on Logic Emerging IoT Adiabatic Techniques …

237

can be achieved by using adiabatic circuits with frequencies ranging from tens of megahertz to hundreds of megahertz. Furthermore, adiabatic logic is better able to withstand attacks from power-based side channels, which makes it excellent for IoT devices with limited resources. Some recent advances in power-clock, performance constraints, and project automation are summarised in this study. The application of adiabatic logic in wireless powered systems has been shown.

References 1. Caro F, Sadr R (2019) The internet of things (IoT) in retail: bridging supply and demand. Bus Horiz 62(1):47–54 2. Chen J, Ran X (2019) Deep learning with edge computing: a review. Proc IEEE 107(8):1655– 1674 3. Madushanki AAR, Halgamuge MN, Wirasagoda WAHS, Ali S (2019) Adoption of the internet of things (IoT) in agriculture and smart farming towards urban greening: a review 4. Networking CV (2016) Cisco global cloud index: forecast and methodology, 2015–2020. White paper. Cisco Public, San Jose, 2016 5. Yu W, Liang F, He X, Hatcher WG, Lu C, Lin J, Yang X (2019) A survey on the edge computing for the internet of things. IEEE Access 6:6900–6919 6. Azar J, Makhoul A, Barhamgi M, Couturier R (2019) An energy efficient IoT data compression approach for edge machine learning. Future Gener Comput Syst 96:168–175 7. Moon Y, Jeong D-K (1996) An efficient charge recovery logic circuit. IEEE J Solid-State Circuits 31(4) 8. Rajesh A, Raju BL, Reddy KCK (2014) Reduction of power dissipation & parameter variation in VlSI circuits for SOC. Int J Rev Electron Commun Eng (IJRECE) 2(3) 9. Landauer R (1961) Irreversibility and heat generation in the computing process. IBM J Res Dev 5(3):183–191 10. Frank MP (2005) Introduction to reversible computing: motivation, progress, and challenges. In: Conferences on computing frontiers, pp 385–390 11. Moon Y, Jeong D-K (1996) An efficient charge recovery logic circuit. IEEE J Solid-State Circuits 31(4):514–522 12. Gong C-SA, Shiue M-T, Hong C-T, Yao K-W (2008) Analysis and design of an efficient irreversible energy recovery logic in 0.18-m cmos. IEEE Trans Circuits Syst I Regul Pap 55(9):2595–2607 13. Maksimovic D, Oklobdzija V, Nikolic B, Current KW (1977) Clocked cmos adiabatic logic with integrated single-phase power-clock supply: experimental results. In: International symposium on low power electronics and design, pp 323–327 14. Celis-Cordova R et al (2019) Design of a 16-bit adiabatic microprocessor. In: IEEE international conference on rebooting computing, 2019, pp 1–4 15. Wan T, Karimi Y, Stanacevi´c M, Salman E (2017) Perspective paper—can AC computing be an alternative for wirelessly powered iot devices? IEEE Embed Syst Lett 9(1):13–16 16. Wan T, Karimi Y, Stanacevic M, Salman E (2017) Energy efficient AC computing methodology for wirelessly powered IoT devices. In IEEE Int Symp Circuits Syst 17. Wan T, Karimi Y, Stana´cevi´c M, Salman E (2019) Ac computing methodology for RF-powered IoT devices. IEEE Trans Very Large-Scale Integr Syst 27(5):1017–1028 18. Huang Y, Wan T, Salman E, Stanacevic M (2019) Signal shaping at interface of wireless power harvesting and ac computational logic. In: IEEE international symposium on circuits and systems, May 2019 19. Moon Y, Jeong D-K (1996) An efficient charge recovery logic circuit. Solid-State Circuits IEEE J 31(4):514–522

238

T. Vijayalakshmi and J. Selvakumar

20. Wan T, Salman E, Stanacevic M (2016) A new circuit design framework for IoT devices: charge recycling with wireless power harvesting. In: IEEE international symposium on circuits and systems, May 2016 21. Avital M et al (2015) DPA-secured quasi-adiabatic logic (SQAL) for low-power passive RFID tags employing S-boxes. IEEE Trans Circuits Syst I Regul Pap 62(1):149–156 22. Lu S, Zhang Z, Papaefthymiou M (2015) 1.32 GHz high-throughput charge-recovery AES core with resistance to DPA attacks. In: IEEE symposium on VLSI circuits, June 2015, pp C246–C247 23. Kumar SD, Thapliyal H, Mohammad A (2018) Finsal: finfetbased secure adiabatic logic for energy-efficient and DPA resistant IoT devices. IEEE Trans Comput-Aided Des Integr Circ Syst 37(1):110–122 24. Wan T, Salman E (2018) Ultra low power simon core for lightweight encryption. In: IEEE International symposium on circuits and systems, May 2018 25. Leff HS, Rex AF (eds) (2003) Maxwell’s Demon 2: entropy, classical and quantum information, computing. Institute of Physics Publishing 26. Schlaffer A, Nossek JA, Is there a connection between adiabatic switching and reversible computing? Institute for Network Theory and Circuit Design, Munich University of Technology, http://citeseer.nj.nec.com/schlaffer97is.html 27. Frank MP (1999) Reversibility for efficient computing, manuscript based on Ph.D. thesis, Dec. 1999, http://www.cise.ufl.edu/~mpf/-manuscript, §7.6.4, pp 197–199 28. Zurek WH (2002) Decoherence, einselection, and the quantum origins of the classical. Preprint http://arxiv.org/abs/quant-ph/-0105127 29. Hall JS (1992) An electroid switching model for reversible computer architectures. In: PhysComp ’92 (ibid. [2]), pp 237–247 30. Younis SG, Knight TF Jr (1994) Asymptotically zero energy split-level charge recovery logic. In: International workshop on low power design, pp 177–182. http://www.cise.ufl.edu/~mpf/scrl94.pdf 31. Younis SG (1994) Asymptotically zero energy computing using split-level charge recovery logic. Ph.D. thesis, MIT EECS Dept. http://www.cise.ufl.edu/~mpf/younis-phd.ps 32. Vieri C, Ammer MJ, Amory Wakefield L (1998) “Johnny” Svensson, William Athas, and tom knight “designing reversible memory,” unconventional models of computation, Springer, pp 386–405

A Critical Review of Agri-Food Supply Management with Traceability and Transparency Using Blockchain Technology Sanket Araballi and P. Devaki

Abstract Agriculture is the backbone of our society; hence, the Indian economy is heavily reliant on farmers. Farmers are in charge of crop cultivation and account for around 51% of all agricultural production. Despite their contributions, individuals do not realise the benefits or earn sufficient profit for a variety of reasons, including a lack of understanding and supply chain management inefficiency. Smart technologies that require supply chain management models are used to tackle these difficulties. This aids in the financial transaction’s monitoring at each stage. Blockchain technology has recently emerged as a transparent supply chain management platform. The goal of this study is to show how various supply chain management systems can track their transparency. We show the various problems with the current system in this review for a traceable transaction that can help farmers in tracing the financial transaction. Furthermore, AI (Artificial Intelligence) is suggested for future research direction. Keywords Transaction traceability · Agri-Food

1 Introduction Agriculture is the world’s backbone, having a direct impact on a variety of fields, ultimately influencing the main purpose of human existence. Agriculture is important for a country’s economy, as well as providing security, and nutritional value, and promoting population health. Agriculture carries several risks, such as changing weather patterns from season to season. Agriculture products’ costs fluctuate dramatically due to soil degradation, weeds, pests, and non-sustainable crops, which affect yield and Climate change. Figure 1 shows the Agri-Food Supply management and explained later. S. Araballi (B) · P. Devaki Department of Information Science and Engineering, NIE, Mysuru, Karnataka, India e-mail: [email protected] P. Devaki e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_18

239

240

S. Araballi and P. Devaki

Fig. 1 Simplified version of the Agri-Food supply chain management process

The above figure i.e. Figure 1 shows the simple version of the supply chain management process, the roles of the actors involved are stated below [1]: (A) Provider: is responsible for providing raw materials like seeds, nutrients, pesticides chemical fertilizers etc. (B) Producer: the farmer is responsible for actions starting from the planting of seedlings to harvesting the crops. (C) Processor: this actor is involved in the simple processing of complex tasks. (D) Distributor: The Distributor is accountable to transfer the end product of the processor from the processors’ stage to the site of retailers. (E) Retailer: The retailer is accountable to sell the products, by representing in provisional stores to supermarkets. (F) Consumer: This actor is the end output in the chain, who act laterally through the entire process, the authorities deliver standard measures, regulations, laws, rules and policies that are added and conform with. Furthermore, there are various phenomena and process which is carried out for ideal farming [2–4]; these are explained as follows: i.

ii.

iii.

iv.

Raw materials: The producers and the providers stock blockchain technology with sales information and purchase history of the raw materials supplied, which include technical data of the products and their price. To automatize the process smart tags are used. Planting: The producers stock up the blockchain data around the planting process e.g.; the total calculation of seeds used; the sensors automate this data entry process. By connecting various weights for smart contracts that spontaneously explode, that create accounts when an anomaly is sensed (more seeds need to be registered unlike those that are purchased). Growing: The sensors are planted at various points that separately store the blockchain data to keep track of the plants grown and the environment considered. Smart contracts that track records when an anomaly is detected, sensor values are said to be out of threshold. Farming: Farmers are responsible for storing the blockchain data at every stage in the system model, like the amount of input applicable. Sensors are capable

A Critical Review of Agri-Food Supply Management with Traceability …

241

of automating the entry process for data generated, which involves chemical sensors and multisensory systems that include smart contracts which automatically detect fire henceforth, create records when an anomaly is detected, and sensor values achieved outside of the threshold. v. Harvesting: the farmers are responsible to store blockchain data about harvesting. Sensors can automate the data entry connected through scales. Smart contracts are responsible for autonomous fire, henceforth to certify the process right from seed harvesting that is compliant with certain regulations. vi. Delivery to the processor: farmers are responsible for transferring the ownership to the local distributors, via the blockchain. Sensors and smart contracts responsible to automate the process create records when anomalies are sensed in the delivery phase. vii. Processing: Consider the simple scenario of packaging a processor, the latter stores the blockchain technology information approximately by the amount received by the product or distributors. The packed amount of the product is misplaced during the processing phase. Smart contracts automate the data or create a new process when an anomaly is detected during the delivery phase. viii. Retailing: retailers are responsible to store the blockchain data about the amount received as a product from distributors. At regular intervals sensors autonomously save the blockchain information of the status details of the retail environment. Note: smart contacts asynchronously fire, for certain records when an anomaly is detected. ix. Consuming: retail shopkeepers store the blockchain details of the products sold, as these customers validate the history of a product before purchasing it. Smart tags are connected along with each package for consumers the regain the end system product. Traceability In Agriculture, Transaction traceability anticipates a significant volume of data created by the supply chain. Employees in the early track and traceability system take track of data in the field, which is later manually recorded to a handbook or into a computer model [5]. This approach poses risks that as erroneous data recording and ineffective resource usage. In the last decade, we saw a rapid increase in communication skills to automate processes and products, resulting in the Internet of Things (IoT) framework [6]. Figure 2 shows the agriculture traceability with respect to IoT. The quick expansion of (the Internet of things) and sensor technology favours data collection methods that are both fast and reliable. This system combines blockchain technology for identifying the product, analysing the ingredient, transport, and data recording. This strategy integrates blockchain technology in identifying the product, analysing the ingredient, shipping, and data gathering, along with overall system integration. Various supply chains take advantage of technologies such as barcodes, QR codes, and RFID (WSNs). RFID systems protect information and data for farmers, distributors, merchants, and consumer management solutions for agri-foods [7]. To trace and monitor the “farm to fork route,” RFID technology aids in the management

242

S. Araballi and P. Devaki

Fig. 2 Agriculture traceability

of the agri-food supply chain. When a problem with food safety emerges, the source of the problem and, as a result, the solution is discovered quickly. Blockchain technology incorporates storage of the transactional data and then saves and simplifies programs to implement this. To enable this application so that it is distributed [7]. In the article founder of Ethereum said that decentralization refers to three levels of (de) centralization. 1. Logical (de) centralization; 2. Political (de) centralization; 3. Architectural (de) centralization [8]. To this day many frameworks utilize the certification process that ensures not deliver complete distinguishability to the adjacent nodes of the network. To recognize the essential problem in supply chain management applications, such as product supply chains, in a setting where “smart contracts” can be used in a promiscuous manner. Nick Szabo coined the phrase “smart contract” in 1997. His main idea was smart contracts similar to contracts in the real world with a slight difference from originality. These are digital, small programs applied and kept in blockchains. To feature this break-free logical code into it. Smart contracts are responsible to get a few attributes, they are absolute and spread across blockchain technology. This absolute nature ensures that we cannot meddle with the code of the contract. For being distributed this secure evaluation of smart contracts are similar to blockchain applications like that in agri-food supply chains and the health sector [9]. When an operation is scanned and found thoroughly in a complete digitalized way. The exact operation is confined to that block and added to the supply chain. Once the agreement is executed at the chain. A statement issued by a variety of information relevant to a blockchain is added [10]. Each consumer in the network holds a duplicate of the smart contract resulting in the following: 1. History of all smart contracts, 2. History of all transactions; 3. The current state of all smart contracts [11]. Ethereum is implemented for smart contract sustenance. As a result, numerous programming languages enable software coding processes in conjunction with blockchain technology.

A Critical Review of Agri-Food Supply Management with Traceability …

243

A widely used tool to support Ethereum’s solidity programming language, a Turing-complete language that establishes the sequence of particular rules that govern how a programme operates and executes [11].

1.1 Motivation and Contribution of Research Work In general, supply chain management is focused on a traceability model that makes use of blockchain’s features, in which all stakeholders access each operation and data connected to a particular product. The main objective is to cover the entire data collection and management process. For each transaction involving participants in the agriculture supply chain. This entire system is known as a “farm-to-fork” approach since it allows to monitor, track, and trace the value of agricultural products. In comparison to centralized methods, Tian’s solution has both benefits and drawbacks. Thus, motivated by the problem faced by farmers for transparency, this research review conducts a critical review of supply chain management with an aspect of transaction traceability and transparency.

2 Related Work The Agri-food supply chain system includes several levels of transactions, each level has a dissimilar set of terms and conditions [12]. Various systems have different features and the function has been working together with supply-chain, which includes food processing, transportation, storage and distributors. Following that, an examination of the electronic traceability systems for agricultural products that describe the condition described above was conducted. As blockchain technology matures and enters with a variety of applications, research is carried out extensively that concentrate using of DTL for agriculture and traceability systems. Table 1 shows the comparison of various mechanisms based on different parameters like traceability, Cost, Inventory Management and so on.

2.1 Traceability As a supply chain, blockchain allows users to track agri-food items among stakeholders. Feng et al. [1] have analyzed the traceability issue, blockchain addresses on coordinating, verifying, linking and recording transactions. Blockchain lets trace the products from farms to consumers, [2, 12]. Chan et al. [5] proposed an Information asymmetry which is addressed by an Agri-food supply chain management software built on blockchain for traceability and transparency, allowing information to be shared with supply chain stakeholders. Stakeholders can audit transactions on

244

S. Araballi and P. Devaki

Table 1 Mechanism comparison Review basis

Significance

Traceability

As a supply chain, blockchain allows users to track [1, 2, 5, 7–9, 12] agri-food items among stakeholders that have analyzed the traceability issue, blockchain addresses coordinating, verifying, linking and recording transactions. Blockchain lets trace the products from farms to consumers

Papers

Efficiency

Blockchain increases the performance of the business through minimal effort

[5, 10, 11, 13, 14]

Privacy

To entail business matters confidential blockchain ensures security property and anonymity to maintain relationships undisclosed

[15–18]

Cost

Blockchain technology helps reduce transaction fees and ease fair prices throughout the chain

[5, 8, 15, 19–22]

Inventory management The procedure of observing the transfer of goods [23–25] initiated from production houses to warehouses and transferring these provisions to retail shops is a primary aspect of supply chain management

the blockchain, in addition to providing transparency [6]. Because the blockchain is absolute, the data in it cannot be modified with [7–9]. The transaction history can be viewed and copied by any valid user [8]. They have discussed using the blockchain in food authenticity, which ensures that no changes were made. When it was in the hands of a specific party, a variety of things could happen. As a result, blockchains could be used to combat food theft and improve traceability. However, data collection is required for traceability. Food traceability is complicated by a lack of records [8]. Without telling blockchain consumers, products can be intentionally harmed [2]. Furthermore, ensuring the validity, confidentiality, and integrity of raw input data is problematic [1]. The accuracy of data collected by sensors or by people is not guaranteed. Besides, these analytical methods are unable to screen various food product parameters, particularly environmental parameters. Third parties, like the government, tracking the blockchain network to solve the issue of data manipulation.

2.2 Efficiency Blockchain increases the performance of the business through minimal effort. Leng et al. [11] and Hasan et al. [12] suggested that blockchain technology increases the complete efficiency, throughput, and reliability of relevant frameworks, this eases the growth of the business. Blockchain is responsible for digitizing the product and certifications that provide real-time information about food products while reducing the time it takes to trace from a week to a few seconds. This functionality enables tracing the contaminated products based on plant or animal disease outbreaks [5]. Human

A Critical Review of Agri-Food Supply Management with Traceability …

245

intervention is reduced by trust and self-organizing nature while integrating IoT devices to improve supply chain efficiency [13, 14]. For instance, uses blockchain to boost security independence which exhibits agricultural robotic swarm operations. They leverage blockchain technology to enable the information recorded for autonomic payment transfer. Blockchain technology spontaneously along the smart contracts could implement and assess transactions to stop. Further, added mutilation with cautions Applicants assess declarations carried out and inform when relevant parties give in to the condition such as quality, timing, and quantity. Smart contracts deal with mountable and easy business at a cheap price, to enhance the complete performance including the manufacturing services.

2.3 Privacy To entail business matters confidential blockchain ensures security property and anonymity to maintain relationships undisclosed. Zhao et al. [15] stated blockchain technology enhances the confidentiality of the enterprise. Blockchain ensures features like anonymity, traceability, reliability, security and transaction visibility to ensure the correctness of agri-food products. Blockchain is associated with more privacy issues, in comparison with benefits that were briefly mentioned. Privacy can be comprised of permanent information to ensure data visibility transparency. Zhao et al. [15] have discussed many issues on privacy leakage and various efforts to ensure privacy. Consider, the example of stakeholders who may want to keep their information confidential to ensure business competitivity. Blockchain technology which with inhibits private or permissioned blockchain is preferred over a public blockchain. A challenge encountered in blockchain technology implementation is the accurate selection of the above technology. The privacy management and transactional privacy cannot be ensured because all participants are not competitors because all applicants in the network access the information [16–18]. There exists no single technique that hides all parties’ involvement, as well as the transaction amount at the same pace. One of the drawbacks of blockchain technology is ensuring Privacy. These issues need to be considered for the management of data, essentially data ownership and data retention [16].

2.4 Cost Blockchain technology helps reduce transaction fees and ease fair prices throughout the chain as suggested by Chan et al. [5]. Another challenge associated with blockchain technology is the cost of computing and expurgating the equipment for the smooth running of the system to adopt a blockchain. Adopting blockchain technology requires several transformations and variations in the organization that are expensive [15] and time-consuming. The cost of blockchain is a hindrance to

246

S. Araballi and P. Devaki

adopting blockchain technology [19]. Many countries face complications since the high degree of computation essential [15]. Hence, [19] stated that blockchain technology is “SME” friendly. In the early stages, once the technology is advancing the high level of uncertainty and market volatility may result in challenges of adoption, because of the old mindset of many stakeholders as well as they may not be mindful of technology and its benefits (e.g., [5–7]) The ideology of blockchain exists that skills employees and organizations. Further satisfying business solutions are limited. Blockchain technology may impose risk on the public to ensure the volatility of cryptocurrencies [5]. The stakeholders adopt a partial education and necessary skills to adapt to blockchain technology [15]. The training framework still gives inadequate results [5]. Cost is sustainable in blockchain technology [19]. The adoption requirements are altered based on different phases of the chain [7, 12]. Hasan et al. [11] suggest minimal changes in the blockchain as a substitute for adopting a new one. Other technologies are most suitable in various cases [7, 12, 14]. Enhancing the knowledge of users’ understanding of blockchain improvises adoption nature [19]. The government plays a key role in minimizing the hindrance caused for adoption by setting different examples [5, 12]. To maintain the capability of businesses for sustaining the blockchain improvises the social, and economic environmental characteristics [5]. Blockchain aids in supervision that assists to abolish the adulteration of food, improvise food safety concerns enhancing the value and solving tough problems, and hence enhancing sustenance (e.g., [3, 8]). Blockchain technology helps allocation of resources [3], to increase the demand–supply and predict quality [4, 7]. Good supervision minimizes the loss and product waste [4], In addition, this provides support for emission reduction [5]. Water management uses blockchain technology to record water quality data to reach sustainability [5–7]. Economically blockchain can improvise the living condition of poor people in developing countries [18] and necessitate food security, (by giving food coupons [10]). Consumers are impacted more by environmental and legal work conditions [8, 20, 21]. Loop [22] stated that blockchain technology acts as a tool for keeping track of social and environmental issues blockchain technology can be used to influence environmental, social, and economic legislation, as well as reduce corruption. Misconduct is held accountable outside the organisation.

2.5 Inventory Management The procedure of observing the transfer of goods initiated from production houses to warehouses and transferring these provisions to retail shops is a primary aspect of supply chain management. This technique isn’t worth the demand forecasting along with inventory management and both are interlinked. If the provider is not able to satisfy the need of customers, the company’s efforts are not utilized and the customers are not satisfied. Consecutively if the supply surpasses the demand, efficient raw materials are depleted for not giving extra profit because there is no possibility to trade additional inventory. High inventory levels accordingly, for various reasons

A Critical Review of Agri-Food Supply Management with Traceability …

247

that confine organizations for profit [23]. Further, the money lent rates increase the expenses of the inventory storage increases well. However, to calculate the supply volume before designing working plans, ML [24] techniques Such dangers should be avoided. Machine learning and data analytics approaches can help with labour management, inventory system automation, vehicle deployment plan optimization, and supply chain decisions that require agility. The ability to accurately estimate demand is crucial for supply chain planning and management [25]. While demand patterns for food remain stable, they demand ambiguity for leisure products like clothes and technology. Blockchain Implementation Challenges significantly.

3 Blockchain Implementation Challenges From the above blockchain technology has gained wide importance and great extension through supply chain management. Blockchain technology and DLT solve most of the prevailing issues, however, these aren’t a solution, to consider the datum of the agri-supply chain which is subtle in terms of criticism and, in many scenarios, they function under difficult pressure. Parallelly to this blockchain technology must also consider various social-technical and financial effects in the agri-supply chain sector or its undone rules and unbreakable relations that aren’t considered. In a further elusive phrasing, the agriculture sector chain is a multifaceted task relevant to blockchain adoption for the following issues. 1. Stakeholders have low-level technical knowledge. 2. The products in the blockchain transform in their entirety. 3. The procedure of role and business allows a huge number of stakeholders that are involved is diverse; 4. The food supply chain is dispersed geographically spread across various continents throughout the globe, it is a fact of the matter for interoperation and deployment across various obstacles. Issues relevant to information management specifically for information ownership and data retention in blockchain technology are essential and needed to be cautiously measured [3]. Blockchains are immutable collections, the data accuracy introduced by the sensors or manually or by people is not guaranteed. Hence, if a sensor breakdown, then the data through the blockchain is inaccurate [5]. These are challenges of monitoring, integrating, and evaluating a certain category of information along the supply chain, in general. Environmental data, for example, is difficult to navigate and analyze using objective methods [8]. Another problem arising here is through the collaboration and continuous integration of blockchain along with legacy traceability problems, to preserve blockchains parallelly through a distributed nature like DLT. The scalability of this transaction system sustains an open challenge which serves as the context for massive executions and deployment [3]. Thus, the proper selection of

248

S. Araballi and P. Devaki

(public, private, permissioned) that implement the blockchain technology is considered an issue. Nevertheless, this benchmark of choice for which the performance is not discussed always. This is because agri-food supply chains involve multiple parties in the transaction’s completion, which necessitates the use of a public blockchain for a significant period and compute capacity. In this regard, the sort of blockchain being explored should be carefully evaluated. This might affect the traceability of the performance of the system. From a financial aspect blockchain increases Because of the transparency of the agricultural food supply chain and consumer trust, utilization is based on a large quantity of energy and financial cost [1]. Companies are required to invest a significant amount of money and time. The amount of time it takes to train staff and get essential material. The time and money invested in traceability costs exceed the entire cost of the product in the short and long term, necessitating a cost–benefit analysis. Given the obstacles ahead, this is an intriguing element to explore. Blockchain applications in the agricultural supply chain were investigated. This research aids in the discovery of 18 food traceability boundary conditions. Five of them have worked with blockchain directly. A vast number of concerns, according to the authors, are related to regulatory requirements for the internal supply chain and industrial processes that necessitate organizational adjustments to offer sustenance while maximizing the aid of traceability. Access to blockchain technology, governance and sustainability, policy and regulation, and the key role of government as a leading example in digitalizing public administration are all important factors. These argue with investment in technology and education that is analysed for further production and validate proof by taking care of the benefits to further yield and demonstrate the proof given the advantages of new blockchain technology. The previous sub-section gives an outline of the challenges of blockchain technology for the probable integration of the agriculture sector. Considering these challenges future research is to tackle and overcome these issues. Safety must be ensured on the blockchain technology, parallelly making it more flexible. Blockchain technology may also be considered an established method concerning financial technology and cryptocurrencies, as long as the supply chain is still deemed to be in its initial stage. As a result, in addition to technical difficulties and tasks, the additional debate will focus on the necessary legislation and policies for incorporating blockchain technology which is a brittle and sensitive sector as that like agriculture. Henceforth, the technologies which are developing faster involve AI by combining with DLT which leads to a smart agriculture platform, where services are rendered by various services, components and stakeholders which are interrelated. Smart agriculture has various benefits despite enhancing traceability systems, and more effective yield by using big data and machine learning algorithms. To carry out this study, needed to be driven toward a more realistic method for making platforms for pilot applications lately. For widening the borders of DLT in combination with big data, AI and machine learning approaches, by making a smart and safe agriculture sector.

A Critical Review of Agri-Food Supply Management with Traceability …

249

4 Conclusion The primary aim of this survey is to carry out a thorough literature review, regarding traceability and blockchain technology which is applied to the agriculture sector. Traceability is a technique that extensively studied these years. Many regulations, directives and laws throughout the world by considering the traceability issues of agri-food products that are set up. Also, blockchain technology is a matter of widespread research. Recently, some research study is carried out on the implanting the blockchain on agriculture traceability systems through their arrival. Simultaneously an enhanced trend of start-ups and pilot applications. The study, carried out through the entire context of this paper represented that blockchain technology seems appealing, but a few limits needless to be taken care of and addressed, like regulations, relationships amongst stakeholders, data ownership, scalability etc. For a good understanding of blockchain technology, the possibility to devise executions, researchers, and developers have many advantages from a widespread valuation system. Towards the conclusion of the implementation, the novel model is considered to be effective, this model leads to authorising these concerns. Minimize the cost, the risk is reduced, minimize the time taken, and enhance trust and transparency. Stakeholders who entail in adopting an innovative technique of functioning, once these are considered that our proposed method is welcoming, enhances efficiency and adds worth. Considering the overhead scenario, we can consolidate the new blockchain technologies through the primitive phase of agriculture as a widespread task that is analysed step-by-step, we can effectively engage the stakeholders who are directly affected by the supply chain.

References 1. Feng H, Wang X, Duan Y, Zhang J, Zhang X (2020) Applying blockchain technology to improve Agri-food traceability: a review of development methods, benefits and challenges. J Cleaner Prod 260. Art. no. 121031 2. Kshetri N (2018) 1 blockchain’s roles in meeting key supply chain management objectives. Int J Inf Manage 39:8089 3. Thakur M, Donnelly KA-M (2010) Modeling traceability information in soybean value chains. J Food Eng 99(1):98105 4. Badia-Melis R, Mishra P, Ruiz-García L (2015) Food traceability: New trends and recent advances. A review. Food Control 57:393401 5. Chan KY, Abdullah J, Shahid A (2019) A framework for traceable and transparent supply chain management for Agri-food sector in Malaysia using blockchain technology. Int J Adv Comput Sci Appl 10(11):149–156 6. Caro MP, Ali MS, Vecchio M, Giaffreda R (2018) Blockchain-based traceability in Agri-food supply chain management: a practical implementation. In: Proceedings of IoT vertical topical summit agriculture Tuscany (IOT Tuscany), May 2018, pp 1–4 7. Watanabe H, Fujimura S, Nakadaira A, Miyazaki Y, Akutsu A, Kishigami JJ (2015) Blockchain contract: a complete consensus using blockchain. In: Proceedings of IEEE 4th global conference consumer electronics (GCCE), Oct 2015, pp 577–578

250

S. Araballi and P. Devaki

8. Queiroz MM, Telles R, Bonilla SH (2019) Blockchain and supply chain management integration: a systematic review of the literature. Supply Chain Manage Int J 25(2):241–254 9. Liang G, Weller SR, Luo F, Zhao J, Dong ZY (2019) Distributed blockchain-based data protection framework for modern power systems against cyber attacks. IEEE Trans Smart Grid 10(3):3162–3173 10. Leng K, Bi Y, Jing L, Fu H-C, Van Nieuwenhuyse I (2018) Research on agricultural supply chain system with double chain architecture based on blockchain technology. Future Gener Comput Syst 86:641–649 11. Hasan H, AlHadhrami E, AlDhaheri A, Salah K, Jayaraman R (2019) Smart contract-based approach for efficient shipment management. Comput Ind Eng 136:149–159 12. Behnke K, Janssen MFWHA (2020) Boundary conditions for traceability in food supply chains using blockchain technology. Int J Inf Manage 52. Art. no. 101969 13. Li Z, Wang WM, Liu G, Liu L, He J, Huang GQ (2018) Toward open manufacturing: a cross-enterprises knowledge and services exchange framework based on blockchain and edge computing. Ind Manage Data Syst 118(1):303–320 14. Ferrer EC (2018) The blockchain: a new framework for robotic swarm systems. In: Proceedings of future technologies conference Cham, Switzerland: Springer, 2018, pp 1037–1058 15. Zhao G, Liu S, Lopez C, Lu H, Elgueta S, Chen H, Boshkoska BM (2019) Blockchain technology in Agri-food value chain management: A synthesis of applications, challenges and future research directions. Comput Ind 109:83–99 16. Demestichas K, Peppes N, Alexakis T, Adamopoulou E (2020) Blockchain in agriculture traceability systems: a review. Appl Sci 10(12):1–22 17. Kosba A, Miller A, Shi E, Wen Z, Papamanthou C (2016) Hawk: the blockchain model of cryptography and privacy-preserving smart contracts. In: Proceeding of IEEE symposium security privacy (SP), May 2016, pp 839–858 18. Lu Q, Xu X (2017) Adaptable blockchain-based systems: a case study for product traceability. IEEE Softw 34(6):21–27 19. Duan J, Zhang C, Gong Y, Brown S, Li Z (2020) A content-analysis based literature review in blockchain adoption within food supply chain. Int J Environ Res Public Health 17(5):1784 20. Lee HL, Mendelson H, Rammohan S, Srivastava A (2017) Technology in agribusiness: opportunities to drive value. Stanford Graduate School Bus., CA, USA, White Paper, Aug. 2017. [Online]. Available: https://www.gsb.stanford.edu/faculty-research/publications/technologyag ribusiness-opportunities-drive-value 21. Loop P (2016) Blockchain: the next evolution of supply chains. Mater Handling Logistics 71(10):22–24 22. New S (2010) The transparent supply chain. Harvard Bus Rev 88(10):76–82 23. Goldratt EM, Cox J (2016) The goal: a process of ongoing improvement. Routledge, Abingdon, UK 24. Abolghasemi M, Beh E, Tarr G, Gerlach R (2020) Demand forecasting in supply chain: the impact of demand volatility in the presence of promotion. Comput Ind Eng 142:106380. [CrossRef] 25. Zhou H, Benton W Jr (2007) Supply chain practice and information sharing. J Oper Manage 25:1348–1365. [CrossRef]

Face-Anti-spoofing Based on Liveness Detection Shivani Mangal and Khushboo Agarwal

Abstract Many applications, like crossing points, banking, and mobile banking, are now using Face Recognition (FR) systems. The widespread usage of FR systems has heightened concerns about the security of face biometrics against spoofing assaults, in which a picture or video of a valid user’s face is employed to attain unauthorized access to resources or activities. Even though numerous FAS or liveness detection techniques (which identify if a face is live or spoofed at the moment of acquisition) have been developed, the problem remains unsolved because of the complexity of identifying discriminatory and operationally affordable spoof characteristics and approaches. Furthermore, particular facial sections are frequently repetitive or correspond to image clutter, resulting in poor overall performance. This paper proposed a neural network model for face-anti-spoofing which outperforms the other models and shows an accuracy of 0.91%. Keywords Face Anti Spoofing · Face recognition · Convolutional neural network

S. Mangal (B) · K. Agarwal Computer Science and Engineering, Madhav Institute of Technology and Science, Gwalior, India e-mail: [email protected] K. Agarwal e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_19

251

252

S. Mangal and K. Agarwal

1 Introduction Biometrics technology has gained in popularity as a result of the quick expansion of Internet technologies, and it is now extensively used in intelligence protection, criminal proceedings, financial and social stability, clinical training, and other disciplines. The face identification system is more simply accepted by the public than extant biometric identification systems owing to its excellent security, genuineness, and non-contact, and has formed an important research path for academics and industries [1]. The face recognition (FR) technology, on the other hand, is open to malware activity by unauthorized users, posing a serious threat to the system’s integrity. As a result, creating a facial anti-spoofing system with higher identification performance, quick response time, and high robustness is critical [2]. The method of determining whether the recently collected facial picture is from a living human or a deceiving face is known as face anti-spoofing (FAS) detection. FAS research has been particularly engaged in recent times both domestically and overseas, owing to its significant academic significance. Printing, video replay and 3D mask attacks are the most popular spoofing assaults. Real and misleading faces have some variations, which are mostly expressed in image texture data, movement details, and perspective details [3]. We can create several FAS systems to identify the actual and counterfeit faces by taking benefit of these distinctions. FAS identification research has progressed fast in recent years, yielding numerous useful research outcomes. This study will examine the methodology based on deep learning (DL), as well as the technique’s merits and weaknesses, as well as the FAS development trend. With DL’s continued advancement and remarkable performances in the field of FR, an increasing number of investigators have used FAS to investigate more comprehensive techniques for combating face deception. DL, as opposed to the old manual feature extraction (FE) technique, may autonomously learn photos, retrieve more critical and plentiful facial features, and assist in effectively distinguishing real from fake faces. They first suggest a (CNN) [4] to extract features in FAS, which paved the way for a new branch of DL in the field of FAS [5]. The recognition impact was significantly lower than that of conventional approaches because the technologies were not yet established. Furthermore, the superiority of DL in feature extraction prompted a significant amount of research to pursue DL-based FAS. FAS based on DL has progressively advanced through network updates, TL [6], a combination of various characteristics, and domain generality, and has now exceeded the previous technique due to the unwavering dedication and repetitive tries of several researchers [7].

Face-Anti-spoofing Based on Liveness Detection

253

2 Related Work Despite significant developments in facial recognition systems, face spoofing remains a significant risk. Most academic and corporate FR systems can be fooled by the following: an image, a video, a 3D face model of a genuine user; a reverse-engineered face image from the template of a genuine user; a sketch of a genuine user, etc. We present a quick summary of published facial impersonation recognition techniques. CNN has proven superior to alternative learning frameworks in a variety of computer vision tasks. For facial pictures, a distinctive feature representation approach known as HGC-CNN is employed to identify face spoof attacks with color photos. It’s a multi-feature learning system that combines capsule NN and hypergraph regularisation concepts. Capsule NN can incorporate a variety of characteristics, including intensity values, LBP, and picture quality. Hypergraph regularisation can also be employed to learn relationships between samples. The expressive ability of extracted features is improved even more when locality information is included. SVM was utilized in the studies since the new representation is consistent with existing classifiers. The suggested approach outperformed the prior approach on FSA detection with color photos, according to experimental data on the NUAA database and the Multispectral spoofing database [4]. An another approach that combines two CNN streams presented by Yousef Atoum et al. They utilize both the whole-facial image and regions taken from a similar face to differentiate the spoof from live faces, as with most previous methods in face anti-spoofing that only use the entire face to identify presenting attacks. The first CNN streaming is based on the characteristics of patches collected from different face areas. This stream proves to be resistant to all types of presentation attacks, particularly on lower-resolution face photos. The second CNN stream uses the whole facial image to estimate face depth. The outcomes of this CNN’s trials suggest that our depth estimation, especially on higher-resolution images, can produce impressive outcomes [8]. Gene LBPnet, a novel technique for CNN based on LBP for face spoofing detection, is presented by Karuna Grover & Rajesh Mehra. On the NUAA dataset, this methodology outperformed previous stateof-the-art algorithms. Using various assessment parameters, it has been demonstrated that the suggested approach provides excellent accuracy (98%) and a low Equal Error Rate, leading to improved recognition of spoofing attacks and thereby improving system security spoofing attempts [9]. To mutually assess the complexity of face pictures and the rPPG signal of face footage, the suggested system integrates CNN and RNN structures. To discriminate between real and fake faces, the approximated depth and rPPG are combined. They also provide a new FAS database for faces that includes a wide range of lighting, subject, and pose variants. The SiW dataset, which covers more subjects and modifications than previous datasets, is introduced. Lastly, they illustrate the technique’s advantage in the experiment [10]. For face liveness identification, Zahid Akhtar et al. propose seven unique strategies for obtaining exclusionary patches in a facial image. A particular classifier is given the properties of specified discriminative picture patches. For the ultimate categorization of authentic and spoof faces, the categorization outcomes of these regions are pooled

254

S. Mangal and K. Agarwal

using a majority-voting-based scheme. In comparison to prior efforts, experiment outcomes on two publically accessible datasets reveal comparable outcomes [11]. To improve the security level of a FAS system, they introduced a novel model for identifying liveness attack images in this article. The variation between the attributes of actual and false faces is taken into account in the approach. As a result, integrating types of image information improves attack effectiveness greatly when compared to using a single approach [12].

3 Proposed Methodology • • • •

Step 1: Collect the CASIA v2 image dataset which is freely available. Step 2: Cleaning the data and removing the noisy data. Step 3: Identifying and removing noisy images and perform data shuffling. Step 4: Reshaping the data features, and samples and splitting them into training and testing. • Step 5: Passing the data into the training model. • Step 6: Train and test samples (3331, 833) for fake and real images and split into 70% for training and 30% for testing. • Step 7: After completion of training measure performance parameters accuracy, recall and precision (Fig. 1).

Fig. 1 Flow chart of proposed methodology

Face-Anti-spoofing Based on Liveness Detection

255

3.1 Dataset Gathering The suggested method is evaluated using the CASIA v2 picture dataset, which is frequently used to identify image forgery and is freely available. There are 4795 photos in all, with 1701 legitimate and 3274 fake.

3.2 Data Pre-processing The goal of pre-processing is to optimize graphic data by overwhelming unwanted deformities or improving particular graphic properties that are important for subsequent processing and evaluation. (a) Data cleaning is the act of determining and restoring (or eliminating) corrupted or erroneous information from a record set, table, or database. It includes recognizing insufficient, improper, faulty, or redundant data and then updating, changing, or deleting the dirty or imprecise data. (b) Checking Noisy Images: Image noise is a sort of ambient sound that produces erratic changes in image intensity or color details. The image detector and circuits of a scanner or digital camera can make it. Movie coarse and the inevitable impulse noise of an optimal photoelectron can likewise cause image noise. We can check the original and noisy images in the dataset and convert all the images to error analysis for better performance. (c) Data Shuffling: The shuffling strategies try to jumble up data while retaining logical linkages among columns if desired. It rearranges data from data inside a feature (for example, a column in pure flat format) or a collection of attributes randomly (e.g. a set of columns). Figure 2 shows the original and ELA image.

3.3 Model Parameter Figure 3 shows the model parameter and explain is below: (1) Conv2D: Conv2D is a 2-D convolution layer that produces a sequence of results by twisting a convolution kernel with the layers’ data [13]. (2) Max-Pooling: Pooling that chooses the largest component from the section of the feature map encompassed by the filters is known as max pooling. As a consequence, the result of the max-pooling layer would be an FM with the most important characteristics of the previous FM [14]. (3) Dropout Layer: Dropout is a strategy for avoiding overfitting in a model. At every iteration of the training stage, Dropout consists of setting the outbound edges of hidden nodes (Hidden components are made up of neurons) to 0 [15]. (4) Flatten Layer: The process of converting data into a 1D array for usage in the following layer is known as flattening. The CL result is flattened to produce a

256

S. Mangal and K. Agarwal

(a) Original Image

(b) ELA Image Fig. 2 Figure showing the original and ELA image

single long feature representation. It’s also related to a fully-connected layer, which is the definitive classification technique [16]. (5) Dense Layer: A DL in any NN is tightly linked to the layer before it, indicating that each of the layer’s neurons is linked to each of the layer’s neurons. It is the most commonly used layer in ANN. The outcome of the DL is an ‘m’ dimensional array. As a consequence, the layer is typically used to change the dimensionality of the vector. The vector is also subjected to processes such as rotation, scale, and translation by these layers [17].

Face-Anti-spoofing Based on Liveness Detection

257

Fig. 3 Model parameter

The neural network model is sequentially trained. The employed NN model with layers is shown in Fig. 3. The dataset is separated into the training of 70% and testing of 30% images for fake (3331) and Real (833) images. The NN is trained and used the RELU and Sigmoid as the activation function. The first layer of the network is the conv2D layer with 2432 parameters, after that the max-pooling2D layer is employed proceeding again to conv2D and max-pooling layer. The dropout, flatten and dense layers were then employed in a cascade manner. Table 1 shows the hyper parameters of training where the ADAM optimizer is used with 20 epochs for a batch size of 32. Table 1 Hyper parameters of training

Optimizer

ADAM

Loss function

Binary cross-entropy

Metrics

Accuracy

Epochs

20

Batch size

32

Validation split

0.2

Shuffle

True

258

S. Mangal and K. Agarwal

4 Simulation Result 4.1 Performance Matrix Precision and accuracy: The degree to which a measured value is near its true value is known as accuracy. Precision refers to how closely all of the measured values are related. To put it another way, accurateness is the proportion of right categories to total classifications. Recall/Sensitivity: Sensitivity is defined as the proportion of true positives to the whole number of actual positives. Similarly, specificity, also known as the true negative rate, is the proportion of genuine negatives to total negatives [18]. F1-Score: When a model’s accuracy is greater than 90%, it is considered to be accurate, we also include the F1 score as a statistic that provides a better indication of cases that have been wrongly classified. The harmonic mean of precision and recall is employed to compute this. When TP and TN are more significant, accuracy is utilized. When the class distribution is unequal and FP and FN are more important, the F1 score is a better statistic [19]. All of the metrics formulas are as shown below. accuracy =

TP +TN T P + T N + FP + FN

(1)

TP T P + FP

(2)

pr ecision = r ecall = F-scor e =

TP T P + FN

2 1/ pr ecision + 1/r ecall

(4)

TN T N + FP

(5)

TP = r ecall T P + FN

(6)

speci f icit y = sensitivit y =

(3)

4.2 Confusion Matrix In a classification issue, a Confusion Matrix is a tabular representation of prediction outcomes with count values split down by class. It demonstrates how a classification model performs while making predictions, as the name implies. It reveals the types of errors made by the classifier as well as the errors themselves [20]. Better and

Face-Anti-spoofing Based on Liveness Detection

259

worse classification results are represented by the points above and below the line, accordingly. The matrix is shown in Fig. 4. Figure 5 shows the accuracy and loss graphs for the evaluated results. Table 2 shows the comparison of the base and proposed results with the proposed system accuracy of 0.91.

Fig. 4 Confusion matrix showing the true and the predicted label

Fig. 5 Figure showing the loss and accuracy graph

Table 2 Comparison of the base and the proposed results Results

Base Paper

Proposed

Recall

0.81

0.85

Precision

0.86

0.97

F1 Score

0.81

0.91

Accuracy

0.85

0.91

260

S. Mangal and K. Agarwal

Class: Fake Confidence: 99.86

Class: Real Confidence: 99.56 Fig. 6 An example of a resultant image compared with the original image

The precision, recall, and f1-score of the proposed model are 0.97, 0.85, and 0.91. The fake and real confidence of the resultant images is shown below. Figures 6, 7 and 8 shows fake and real confidence images for spoofing techniques with duplicate photographs of people whose original images areas maintained in a database. It means that if an intruder wanted access to the authorized system, he or she may have used these several techniques.

5 Conclusion Face Recognition has become an essential technique for achieving protection as AI has become more widely used in real life. FAS has become a pressing issue in the fight against harmful attacks. The research of face spoofing identification has been continuously monitored and revised, from the starting of manual FE methods based on image texture, image quality, and depth information, to using DL to instantly extract

Face-Anti-spoofing Based on Liveness Detection

Class: Fake Confidence: 97.98

Class: Real Confidence: 99.95 Fig. 7 An example of a resultant image compared with the original image

Class: Fake Confidence: 98.75

Class: Fake Confidence: 98.11 Fig. 8 An example of a resultant image compared with the original image

261

262

S. Mangal and K. Agarwal

features, merged with network up-gradation, feature assimilation, and domain generalization, and the efficiency and effectiveness of identification have now attained a significant state.

References 1. Hashemifard S, Akbari M (2021) A compact deep learning model for face spoofing detection. [Online]. Available: http://arxiv.org/abs/2101.04756 2. Al-Huda Taha N, Hassan TM, Younis MA (2021) Face spoofing detection using deep CNN. Turkish J Comput Math Educ 12(13):4363–4373 3. Ming Z, Visani M, Luqman MM, Burie JC (2020) A survey on anti-spoofing methods for facial recognition with RGB cameras of generic consumer devices. J Imaging (6)12. https://doi.org/ 10.3390/jimaging6120139 4. Liang Y, Hong C, Zhuang W (2021) Face spoof attack detection with hypergraph capsule convolutional neural networks. Int J Comput Intell Syst 14(1):1396–1402. https://doi.org/10. 2991/IJCIS.D.210419.003 5. De Souza GB, Papa JP, Marana AN (2021) Efficient deep learning architectures for face presentation attack detection. 112–118. https://doi.org/10.5753/sibgrapi.est.2020.12992 6. Khalid IA (2020) Transfer learning for ımage classification using tensorflow. Towards Data Sci 7. Zhang M, Zeng K, Wang J (2020) A survey on face anti-spoofing algorithms. J Inf Hiding Priv Prot 2(1):21–34. https://doi.org/10.32604/jihpp.2020.010467 8. Liu Y, Jourabloo A, Liu X (2018) Learning deep models for face anti-spoofing: binary or auxiliary supervision. In: Proceedings of the IEEE computer society conference computter vision pattern recognition, pp 389–398. https://doi.org/10.1109/CVPR.2018.00048 9. Das PK, Hu B, Liu C, Cui K, Ranjan P, Xiong G (2019) A new approach for face anti-spoofing using handcrafted and deep network features. In: Proceeding—IEEE ınternational conferences servey operations and logistics and ınformatics 2019, SOLI 2019, no November, pp 33–38. https://doi.org/10.1109/SOLI48380.2019.8955089 10. Akhtar Z, Foresti GL (2016) Face spoof attack recognition using discriminative ımage patches. J Electr Comput Eng. https://doi.org/10.1155/2016/4721849 11. Atoum Y, Liu Y, Jourabloo A, Liu X (2018) Face anti-spoofing using patch and depth-based CNNs. In: IEEE ınternational joint conference on biometrics (IJCB) 2017, vol 2018-Janua, pp 319–328. https://doi.org/10.1109/BTAS.2017.8272713 12. Grover K, Mehra DR (2019) Face spoofing detection using enhanced local binary pattern. Int J Eng Adv Technol 9(2):3365–3371. https://doi.org/10.35940/ijeat.b3834.129219 13. Pouyanfar S et al (2019) A survey on deep learning. ACM Comput Surv 51(5):1–36. https:// doi.org/10.1145/3234150 14. Masita KL, Hasan AN, Shongwe T (2020) Deep learning in object detection: a review. In: 2020 International conference artificial ıntelligence big data, computing data communication system (icABCD 2020)—proceeding no. August, 2020. https://doi.org/10.1109/icABCD49160.2020. 9183866 15. Manalu BU, Tulus, Efendi S (2020) Deep learning performance in sentiment analysis. In: 2020 4th International conference on electrical telecommunication and computer engineering (ELTICOM 2020)—proceeding, pp 97–102. https://doi.org/10.1109/ELTICOM50775.2020. 9230488 16. Shirahatti AP, A survey of deep learning for sentiment analysis. V(I):1–7 17. Weng W, Zhu X (2021) INet: convolutional networks for biomedical image segmentation. IEEE Access 9:16591–16603. https://doi.org/10.1109/ACCESS.2021.3053408 18. ML (2020) Classification: precision and recall. Machine learning crash course. https://develo pers.google.com/machine-learning/crash-course/classification/precision-and-recall

Face-Anti-spoofing Based on Liveness Detection

263

19. R (2020) ROC Curves. Machine learning crash course. https://developers.google.com/mac hine-learning/crash-course/classification/roc-and-auc 20. Narkhede S (2018) Understanding confusion matrix. Towardsdatascience. https://towardsdatas cience.com/understanding-confusion-matrix-a9ad42dcfd62

PDR Analysis and Network Optimization of Routing Protocols for Edge Networks Archana Ratnaparkhi , Radhika Purandare , Gauri Ghule , Shraddha Habbu , Arti Bang , and Pallavi Deshpande

Abstract AdHoc On-Demand Distance Vector (AODV) is a notable and broadly utilized protocol for MANETs. The Mobile AdHoc Network or MANET, without any infrastructure, is an assortment of remote nodes conveying and communicating over a wireless network. All wireless devices working in AdHoc mode inside range are allowed to have communication with each other in absence of base station. The routers are capable to roam and communicate arbitrarily and organize themselves as per the requirements when the nodes structure themselves into a random topology. Radio signals possess range limitations due to which multihop communication in MANETs is inevitable. The performance of traffic situations utilized in a mobile AdHoc network is responsible for the transmission and gathering of data between source and destination in a MANET.This paper provides a comprehensive comparative analysis of the routing protocols with respect to variation in node configuration. Keywords Adhoc networks · Routing protocols · Packet delivery ratio · Throughput · Dynamic environment

1 Introduction Design of effective routing strategy has been a matter of research for decades. Evolution and advancements in routing strategy using optimally minimal resources is the crux. The designing of a reliable and efficient routing strategy is a very challenging problem due to the limited resources in Mobile Ad-hoc Networks (MANETs). The variables in network conditions such as network partitioning, network size and traffic density dictate the routing strategy to be used in order to make efficient use of the limited resources. The routing protocol is also expected to cater to different levels of QoS required by a diverse range of users and applications. With the advent A. Ratnaparkhi (B) · R. Purandare · G. Ghule · S. Habbu · A. Bang · P. Deshpande Vishwakarma Institute of Information Technology, Pune, India e-mail: [email protected] URL: http://www.viit.ac.in © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_20

265

266

A. Ratnaparkhi et al.

of MANETs, the significant algorithms such as link state algorithm and distance vector have undergone numerous enhancements. In huge Mobile Adhoc Network, traditional link-state and distance vector algorithms do not scale, as periodic updates in routes consume a significant amount of the available bandwidth, consume more power on the side of each node and increase channel contention. To improve on these, numerous routing protocols have been proposed. These protocols can be classified as proactive, reactive and hybrid protocols [1]. Basic functional difference between proactive and On-demand routing protocols lies in the way the routing information is maintained. Routing information is maintained by every node to every other node in the network in case of proactive protocol. On-demand routing protocols maintain information only for the active nodes, thus reducing the overheads in proactive routing protocols. Improving on these two kinds of routing protocols, hybrid routing protocols were proposed. These protocols allow the nodes that lie in close proximity to work in unison and form a backbone to cut down on route discovery overheads, thus increasing stability. These protocols work by maintaining the routes proactively to the nodes nearby and determining routes to distant nodes through an appropriate strategy for route discovery.

1.1 Dynamic Source Routing (DSR) As DSR protocol involve carrying the complete packet from source to destination, it makes the protocol ineffective in large networks, as an increase in diameter of the network will increase the packet overhead. Hence, the overhead may occupy most of the bandwidth in large scale networks that are dynamic in nature. This routing protocol is advantageous over protocols such as LMR, TORA and AODV, including better performance in moderately sized networks.

1.2 Ad-hoc On-demand Distance Vector (AODV) This protocol is advancement over sequence numbering and periodic beaconing procedure of DSDV and employs a route discovery procedure like DSR. There are a couple of major differences between AODV and DSR. In DSR, each packet contains complete routing information, where the packets carry only the destination address in AODV. This implies that AODV has fewer routing overheads than DSR. Also, the address of each node along the route is carried by route replies in DSR. AODV trumps over DSR in its adaptability in highly dynamic networks. Although, large delays may possibly be experienced in route construction, and failure to establish a link may initiate an alternative route discovery, consuming more bandwidth and introducing excess delays with an increase in the size of the network [2, 3].

PDR Analysis and Network Optimization of Routing Protocols . . .

267

2 Related Work In this work, analysis of six protocols such as ZigBee, UWB, Bluetooth, WiFi, WiMax and GSM is carried out. The best and most suitable protocol for an application of intelligent sensor is selected by evaluating it on the basis of the transmission time, the data coding efficiency, the bite error rate, and the power and the energy consumption with network size [4]. The cross layer protocols such as Low Energy Adaptive Clustering, Self Organized TDMA Protocol, Flexible TDMA Protocol, Energy Efficient Fast Forwarding Protocol and D-MAC are analyzed by the authors depending on their performance with the help of NS2 [5]. A survey is done on requirements, challenges in technicalities and already available work on MAC layer protocols related to the support of M2M communication. The issues related to efficiency, scalability and accessibility of fair channel for M2M communication are mentioned in detail. The existing MAC layer protocols and their applicability to M2M communication is also reviewed with the protocols that already developed only for M2M communication. The paper also talks about future research and open problems in it and discussed about in-process standardization efforts [6]. A MAC access algorithm for IEEE 802.15.4 LRWPAN is proposed and implemented by authors based on CSMA/CA protocol, backoff delay, the probability of collision, and retransmission in total networks are the things algorithm focused on reducing. The idea is depended on BE and CW parameter’s dynamically modifying value according to traffic state. The unused slot time of superframe is also manipulated to increase the efficiency rate [7]. Examination of the transmitter of a vehicle hub is started which imparts in-high unique environment and undisturbed network change circumstances. The dissection of regular parcel achievement proportion with throughput of VANET in realistic traffic environment is done. Authors also shown through demonstration that transmitter is affordable for vehicle to vehicle communication. It also shows that in rush hour gridlock situation it shows better results [8]. In order to find the optimal network simulator different open source network simulators are compared depending on parameters like CPU utilization, memory usage, computational time, and scalability by simulating a MANET routing protocol [9–13].

3 Proposed Work Experimentation has been done on NS2 platform. The flowchart as seen in Figs. 1 and 2 starts with initialization of environmental and network parameters, delay, queue and then timing information. In the following experiment we have taken both TCP and UDP AODV/DSR protocol and noted the change in Packet Delivery Ratio with respect to change in the distance between their destination nodes. Here we have taken 4 nodes i.e. source and destination nodes for TCP protocol and same as for UDP AODV/DSR protocol and changed the distance of either one or both destination

268

A. Ratnaparkhi et al.

Fig. 1 Flowchart

nodes. Simulation parameters as indicated in literature have been used and tabulated in Table 1. Case 1: The below NAM window snapshot as seen in Fig. 3 shows that when the distance between UDP which is source one and TCP i.e. source two, nodes is greater than 400, we see that packet loss occurs. However when the distance between the destination node of UDP i.e. node 3 and source node of TCP i.e. node 0 is greater than 400, it results to no packet loss. This is discussed later in result analysis. Case 2: The below snapshot as seen in Fig. 4 of NAM window shows the change in distance of both the destination nodes parallelly by the distance parameter of 50, keeping fixed source nodes. Here both UDP and TCP source nodes are fixed and the destination nodes i.e. node 1 and node 3 are moved parallel to each other. The different results at different points will be discussed later. Case 3: The snapshot as seen in Fig. 5 of the NAM window shows the change in distance of only one destination node is changed by the distance parameter of

PDR Analysis and Network Optimization of Routing Protocols . . .

Fig. 2 Programming steps

269

270 Table 1 Simulation parameters Parameter Channel type MAC protocol Routing protocol Traffic type Simulation time (s) Number of nodes Number of sources Range parameter (m) Queue type Radio propagation models Antenna model

A. Ratnaparkhi et al.

Value Wireless channel IEEE 802.15.4 AODV, DSR CBR/FTP 100 4 2 50 Drop tail Two ray ground Omni-directional antenna

Fig. 3 Case 1: variation of PDR with distance (NAM window)

Fig. 4 Case 2: variation of PDR with distance (NAM window)

PDR Analysis and Network Optimization of Routing Protocols . . .

271

Fig. 5 Case 3: variation of PDR with distance (NAM window)

Fig. 6 Case 4: variation of PDR with distance (NAM window)

50, keeping all nodes at fixed position. Here we have change the position of TCP destination node keeping source of TCP and nodes of UDP fixed. Case 4: The below snapshot of the NAM window shows the change in distance of only one destination node (i.e. UDP) by the distance parameter of 50, keeping all nodes at fixed position (Fig. 6).

4 Result Analysis The following observations were noted during the experiment. • When the distance between UDP and TCP nodes is greater than 400 the PDR is always high i.e. 1. • In case 1, when the distance between source and destination nodes is greater than 300 the PDR decreases to 0 gradually. • In case 2, when the distance between source and destination node of TCP is greater than 300, the UDP protocol will drops packets, as a result PDR decreases abruptly towards 0.

272

A. Ratnaparkhi et al.

• In case 3, when the distance between source and destination node of UDP is increased the PDR increases continuously. • In case 4 as the graph shows, DSR shows some fluctuations for some input values of TCP and UDP compared to AODV.

5 Conclusion In this paper, the desired work is achieved using an NS2 simulator under varying parameters, traffic states and conditions to study and analyze the working and performance of AODV/DSR protocols by taking both TCP and UDP connections. We have observed, noted and calculated the change in Packet Delivery Ratio or PDR by varying the distance between the destination nodes. We have taken 4 nodes for our experiment, source and destination nodes for TCP protocol as well as UDP AODV/DSR protocol. By changing the distance of either one or both nodes, we came across multiple cases. We concluded that the PDR is always high or 1 when the distance is greater than 400. The PDR decreases to zero when the distance is greater than 300. The PDR decreases continuously when the distance between the nodes of UDP is increased. Compared to AODV, DSR shows fluctuations or irregularities for few input values of TCP and UDP. This work can be extended further to include more dynamic movement in nodes and the parameters can be studied. Furthermore mobile adhoc networks which are far more complex can be studied as part of further enhancements. Acknowledgements We are extremely thankful to Department of Electronics and Telecommunication and Research and Development wing for their constant support and encouragement.

References 1. Singh B, Hans R (2015) TCP and UDP based performance analysis of AODV, DSR and DSDV routing protocols under different traffic conditions in mobile adhoc networks. Int J Future Gener Commun Netw 8(2):73–92 2. Singh R, Tripathi P (2021) Performance evaluation of TCP (Transmission Control Protocol) and UDP (User Datagram Protocol) over AODV routing protocols. Int J Modern Sci Eng Technol. Academia.edu. Retrieved November 2, 2022, from https://www.academia.edu/11945555/Performance_Evaluation_of_TCP_Transmission_ Control_Protocol_and_UDP_User_Datagram_Protocol_over_AODV_Routing_Protocols 3. Abolhasan M, Wysocki T, Dutkiewicz E (2004) A review of routing protocols for mobile ad hoc networks. Ad Hoc Netw 2(1):1–22 4. Biradar RV, Patil VC, Sawant SR, Mudholkar RR (2009) Classification and comparison of routing protocols in wireless sensor networks. Spec Issue Ubiquitous Comput Secur Syst 4(2):704– 711 5. Chakkor S, Cheikh EA, Baghouri M, Hajraoui A (2014) Comparative performance analysis of wireless communication protocols for intelligent sensors and their applications. arXiv:1409.6884

PDR Analysis and Network Optimization of Routing Protocols . . .

273

6. Gajjar S, Pradhan SN, Dasgupta K (2012) Performance analysis of cross layer protocols for wireless sensor networks. In: Proceedings of the international conference on advances in computing, communications and informatics, pp 348–354 7. Rajandekar A, Sikdar B (2015) A survey of MAC layer issues and protocols for machine-tomachine communications. IEEE Internet Things J 2(2):175–186 8. Bouazzi I, Bhar J, Atri M (2017) New CSMA/CA prioritisation based on fuzzy control mechanism. Int J Intell Eng Informat 5(3):253–266 9. Karunakar P, Matta J, Singh RP, Kumar OR (2020) Analysis of position based routing Vanet protocols using Ns2 simulator. Int J Innov Technol Exploring Eng (IJITEE) 9(5) 10. Khan AR, Bilal SM, Othman M (2012) A performance comparison of open source network simulators for wireless networks. In: 2012 IEEE international conference on control system, computing and engineering. IEEE, pp 34–38 11. Voitenko I (2022) https://github.com/iuriivoitenko/simpleMANET. GitHub. Retrieved June 13, 2022 12. Shakya S, Pulchowk LN (2020) A novel bi-velocity particle swarm optimization scheme for multicast routing problem. IRO J Sustain Wirel Syst 2:50–58 13. Sathesh A (2020) Artificial intelligence based edge computing framework for optimization of mobile communication. J IoT Soc Mob Analytics Cloud 2(3):160–165

Privacy Threat Reduction Using Modified Multi-line Code Generation Algorithm (MMLCGA) for Cancelable Biometric Technique (CBT) Pramod D. Ganjewar, Sanjeev J. Wagh, and Aarti L. Gilbile

Abstract Nowadays individual’s identity verification is required at many places for authenticity like Government Sector, Private sector, Public Sector etc. Many existing systems are based on either physical documents or on biometric parameters. Biometric system has become a more convenient way for authenticity check. In biometric system, some samples of biometric parameters are collected and stored at the server side for further use. These samples would be used for the verification of the identity of the person. The actual required biometric parameter will be compared with all the existing samples available in the system to match with the registered person. If it matches with one of the existing sample, then that will be authenticated by the system and allowed to perform further operations. But what if, the samples collected by the authority get misused? That needs security from the owner of the system. So, the identity of the samples must be hidden from the operators by some way, which has been focused in this work. A new approach called as Cancelable Biometric Technique (CBT) using Modified Multi-Line Code Generation Algorithm (MMLCGA) is used for storing biometric samples using template. The cancelable method converts the gathered samples and stores it into the system for hiding its original identity. In the verification phase, the system will convert it back to the original sample to be used for identity matching of the person or user. This technique provides more privacy, because of which privacy threats can be reduced. The time and accuracy of the proposed technique is better by 15% and 1.4% respectively, when compared to the existing technique Multi-Line Code Generation Algorithm (MLCGA). Keywords Cancelable biometrics · Revocable · Biometric sensor · Multi-biometrics · Biometric security · Template protection

P. D. Ganjewar (B) · A. L. Gilbile MIT Academy of Engineering, Alandi (D.), Pune, Maharashtra, India e-mail: [email protected] S. J. Wagh Government College of Engineering, Karad, Maharashtra, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_21

275

276

P. D. Ganjewar et al.

1 Introduction The regular authentication system uses tokens like PIN, passwords, cards, IDs etc. These tokens can be misused by anybody, without the presence of real entity, which is the disadvantage of the regular authentication system. The authentication system that uses biometrics is better than the existing one. These systems uses biometric features like fingerprint, thumb impression, IRIS structure, voice, keystrokes, signatures etc., as every user has unique biometric features. There is no need to memorize these features for further use. There is no chance of misplacing or forgetting these features like passwords or other identification mechanisms. Also, it cannot be used without the presence of the real entity. But there is an issue with the storage of these biometric templates. They are generally stored as it is, at any place as per the company guidelines. If the database of these biometric templates gets stolen, then they can be misused, as these are stored in its original form. These templates need to be protected in such a manner that, even after getting the access to such templates, it cannot be misused. The templates need protection in such a way that even after recovery of these templates, the original templates cannot be obtained. In template protection, biometric features need to be arranged in some well- organized way, so that original biometric information is more secure. In case of misuse of the transformed biometric template, new template will be formed by changing the transformation parameters of the biometric template and the new template gets restored on the misused template. The primary goals of this work are as follows: • Non-reversibility: After performing many operations, original templates should not be regenerated from the protected template. • Accuracy: The operations performed for the template protection, should not hamper on the quality of the original template. • Diversity: Same biometric cannot be used by different applications. It will increase the chances of revocability.

2 Related Work Kaur and Khanna [1] presented the concerns regarding the security of the biometric systems. The pseudo-biometric data is procured and kept at safe place and used for verification of the identity of the users. The loss of the privacy can be reduced using pseudo-identity and it allows the revocability in cases of compromise. This method extracts pseudo biometric from its original biometrics and also reduces its size by half of its original size. They analyzed the recognition and protection performance of the proposed method based on the data taken from unimodal and multimodal data generated from biometric data of palm-vein, palm-print, face etc. A new transformed template will be generated based on the distance calculation between the feature points and some random point, using Euclidean distance and the performance also get improved.

Privacy Threat Reduction Using Modified Multi-line Code Generation …

277

Wu et al. [2] addressed template protection using cancelable biometric techniques, for electrocardiograms (ECG) biometrics to decrease the threats. The property of ECG, which makes them difficult to forge or steal is its inherent biological and intrinsic life sign. The feature of ECG to be considered is its R peak. The principle of “Signal Subspace collapse” was used to generate various templates associated with ECG to revoke compromised identity like passwords or PINs. To recognize the identity of the users from their ECGs, the popular multiple signal classification technique was used. In this, without understanding the transformation of the distortion, the identification will be carried out unlike other cancelable techniques. This is the reason that the recovery of original ECGs from their template is very difficult. They have analyzed it using the various experiments on real ECGs of 285 subjects. Jin et al. [3], proposed two, template protection methods using ranking based hashing techniques, which were motivated by cancelable biometric and “Index-ofMax” (IOM) hashing. Two realizations named as Uniformly Random Permutation and Gaussian Random Projections are demonstrated from this hashing notion. This system ensured strong concealment to biometric information, more robust against biometric feature variations. Gomez-Barrero et al. [4], suggested that unlink ability and irreversibility are two international standards for protecting biometric templates. The new method for determining the unlink ability of biometric template was proposed. The efficiency of the technique was verified using four techniques called as Homomorphic Encryption, block re-mapping, biometric salting and Bloom filters. Jin et al. [5], illustrated that, specific strings of binary with fixed length are produced from the kernel learning mechanism for fingerprint information. The first listed procedure was—Minutiae description extraction, second was a method of kernel transformation that was based on kernelized hashing to produce a vector of fixed width, third was Binarization (Segmentation) and fourth was to match. The datasets used for experimentation were FVC2002 and VC2004 to describe the possibility of proposed framework in case of template randomness, accuracy of matching and proficiency. Lee et al. [6], developed a new technique to produce a cancellable fingerprint template without alignment. Based on constant value of rotation and translation a minutia will be calculated. The value of the minutia is based on the orientation of its adjacent local areas. The geometric interconnections between the enrolled models will be retained and transformations will be performed. The transformed templates will be operated to verify the unregistered users of the system. The success and changeability are the criteria used for analysis in the experiments to verify the accuracy of the results. Kelkboom et al. [7], proposed an improved method FRR, in which during enrollment and verification process various biometric instances were used. In this noise, bit errors and HD were decreased. The evaluation of the output of the biometric device was proposed based on Gaussian analytical methods, which specified the number of instances used during the measuring and checking process. The error-detection

278

P. D. Ganjewar et al.

trade off curve was used for the assessment of the machine efficiency, which incorporates FAR and FRR. The testing was done based on the biometric Face Recognition database Grand Challenge V2 and FVC2000. Wong et al. [8] suggested a revocable fingerprint template called multi-line code with low complexity. The proposed multi-line code improved performance. The better network protection offering diversity and revocability using the fingerprint template was created efficiently. However, the proposed method required large storage capacity as the template was stored as a real number string. Jin et al. [9] proposed a protection technique to protect the fingerprint minutiae. Graph-based Hamming Embedding was used to protect against inversion by using a minutiae descriptor, as Minutiae Vicinity Decomposition (MVD). User specific Minutiae Vicinities Collection scheme was then used to improve Randomized MVD. Graph-based Hamming Embedding was used to embed the MVD. Binary template has a strong concealment of the minutiae vicinity that protects the location and orientation of minutiae effectively. Jin et al. [10] presented a fingerprint template protection mechanism to secure the fingerprint minutiae. In order to create a binary template that could be strongly secured against inversion, Randomized Graph-based Hamming Embedding (RGHE) was implemented. This approach, adopted a minutiae descriptor, was called Minutiae Vicinity Decomposition (MVD). The user-specific minutiae vicinities collection scheme then enhanced discrimination of the randomized MVD. Using the graphbased Hamming Embedding technique, it was inserted into a Hamming space. This technique has many advantages like, effectively protects the location and the orientation of the minutiae points, the MVD’s well-preserved discriminability, the templates are quick, and it is a revocable method. However, the MVD features are highly likely to reveal the minutiae vicinity. Li et al. [11] addressed the development of privacy-preserving protocols based on the combination of garbled circuit and homomorphic encryption for securing fingerprint minutiae. The suggested work promised both template and transaction security. To encrypt the template stored on the server, the user’s private key is used. By means of re-encryption, the template may be modified or revoked. Two hybrid protocols were introduced in this paper with the combination of homomorphic encryption and garbled circuit to satisfy the matching minutiae. The key downside of the scheme is that, often less efficiency was incurred. Akdo˘gan et al. [12] suggested a new secure key agreement protocol using biometrics with unordered set of features. The suggested protocol used fingerprints without using any randomly generated key or any random data in the key itself. A thresholdbased quantization process was used to group the minutiae in a predefined neighborhood. Based on the measured similarity score on the common collection of minutiae, the acceptance or rejection decision will be made. Li et al. [13] implemented a security analysis framework by combining information-theoretic and computational security approach. Using decision level fusion, a fingerprint based Multi Biometric Cryptosystem (MBC) was built. This was based on two elements called new security analysis framework for bio-cryptosystemoriented and a realistic MBCD construction based on fingerprints. To further secure

Privacy Threat Reduction Using Modified Multi-line Code Generation …

279

MBCD, hash functions were used in each single biometric trait. However, more storage space was required. Subramaniam et al. [14] proposed a system to generate cryptographic keys using biometrics and conventional cryptographic algorithms. These keys were used to generate shared session key between E-Passport and ES. The best method for authentication was developed by combining conventional cryptography. The proposed approach fulfilled all the security goals and can be used for real time applications to transfer information safely. Barman et al. [15] presented an approach to generate the cryptographic keys from cancellable fingerprint template of sender and receiver. The cancelable fingerprint templates were securely transmitted using a key-based steganography technique. The templates were combined with a concatenation-based feature level fusion technique and their elements are shuffled using the shuffle key to create a unique session key for communication using hash. A revocable key for symmetric cryptography was created using irrevocable fingerprints and the privacy of fingerprints is secured by the Cancellable transformation of fingerprint template. It lacks a session-based cryptographic key, however. Vijayakumar [16] preferred Multimodal Biometric Recognition (MBR) technique over unimodal biometric systems to provide more security. The features like iris, face, finger vein, and palm print were used for getting the highest accuracy to identify the exact person. The use of multiple features from the person improved the accuracy of biometric system. Smys and Wang [17] introduced a CTI system using blockchain to handle the issues of sustainability, scalability, privacy, and reliability. This is capable of measuring organizations contributions, reducing network load, creating a reliable dataset, and collecting CTI data with multiple feeds. This had been tested using various parameters to determine the efficiency of the proposed methodology. Experimental results showed that it can save upto 20% of storage space.

3 Proposed Method The method of cancelable biometric based template safety is proposed in this paper to provide the security and privacy against the problems resulting from the phenomenal use of biometric systems. The cancelable biometric turns a user’s original biometric identity into a pseudo-biometric identity that is used for purposes of storage and matching. In the event of breach, the use of pseudo-identity mitigates privacy threats and facilitates revocability. The proposed system is the modification of the Multi-line Code Generation Algorithm (MLCGA) invented by Wong’s [8], which converts the fingerprint image to the line which contains the minutiae in number format. In the proposed system, to reduce the complexity of the algorithm, Line Code Generation step is modified, in such a way that, on the line passing through the minutiae, instead of drawing multiples circles over the line only one circle is drawn with large radius, as shown in Fig. 2.

280

P. D. Ganjewar et al.

3.1 System Modules The cancelable fingerprint technique is implemented using the modified multi-line code algorithm. The entire fingerprint verification system comprises of three main modules. • Pre-processing: Pre-processing includes processing of scanned images to improve the image quality. The feature extraction is also performed in this step. In this, minutiae points on the fingerprints are detected and marked. These minutiae points are unique for every individual. • Template generation: Fingerprints are captured using scanners in the form of images. Here, the captured fingerprint image is converted to the template. The template generation is performed using modified multi-line code algorithm. Generated template is different than actual fingerprint of user. So this pseudo template is then stored into the database. • Template matching: When an user marks his attendance through biometric system, the user will be searched into the system, and then the user’s captured fingerprint is used for verification purpose. This captured fingerprint is again converted using multi-line code algorithm and it will be matched with previously stored template fingerprint in the database. If the fingerprint is matched then the user will be an authenticated user and he will have permission to enter the premise.

3.2 System Architecture As mentioned, the existing system is based on Wong’s [8] system, called as Cancelable Biometric Technique using Multi-line Code Generation Algorithm (CBTMLCGA). The proposed system is the modification of this to simplify the complexity of multi-line code generation step in the proposed algorithm, and is called as Cancelable Biometric Technique using Modified Multi-line Code Generation Algorithm (CBT-MMLCGA). Also, in the proposed system, different algorithms other than Wong’s system are applied in feature extraction. The system architecture of the proposed system i.e., CBT-MMLCGA is shown in Fig. 1. The system architecture of the proposed system called as CBT-MMLCGA has two main parts: • Enrollment • Verification. Enrolment. When the user comes for the first time into the system, he is new to system. Now he needs to register himself using personal details and enrol himself with his fingerprint. This process is known as Enrolment. Scanned fingerprint of user will be an input to the system. Along with fingerprint, user should enter his personal information like full name, username, date of birth, etc. While enrolment

Privacy Threat Reduction Using Modified Multi-line Code Generation …

281

Fig. 1 System architecture

Fig. 2 a Plane view of Wong’s system, b view of proposed system

using multi-line code algorithm, the fingerprint template is generated and stored into database. Verification. When the user comes for the second time into the system, it means the user is already registered in the system. So his identity needs to be verified with the help of his fingerprint. Only verified users will be allowed to enter into the system. This process is known as Verification. Here, the scanned fingerprint will undergo multiple preprocessing steps like minutiae extraction, contextual filtering, binarization, thinning and minutiae detection. This step will give a number of minutiae points which will be unique for every individual. While verification, scanned fingerprint is again converted into template using multi-line code algorithm and this new template is matched with the previously stored template. The fingerprint verification system has the following four stages:

282

• • • •

P. D. Ganjewar et al.

Extraction of minutiae Generation of multi-line code Permutations in multi-line code Fingerprint matching.

Extraction of Minutiae. The minutiae extraction includes the following five stages: Segmentation: The image contains unwanted part other than the region of interest. Likewise in the fingerprint image, which also contains some unwanted parts that need to be removed by procuring the region of interest using segmentation by excluding the background. Contextual filtering: The contextual filtering is required to improve the image quality. So the quality of fingerprint image will also be improved using this. The noise from the fingerprint image will be removed by conserving the fingerprint structures. The Gaussian filter is one of the option for the contextual filtering. Binarization: In image processing, it is required to convert grey scaled image into black and white. Therefore, using binarization the grey scaled actual image is transformed to a binary image in black and white. Thinning: In image processing, it is required to reduce the ridges to pixels, and that can be done using thinning. Similarly, the ridges of the fingerprint are reduced to one pixel long using image to make the further processes easy. In this work, Zhang–Suen image thinning algorithm is used. Minutiae detection: In the image processing, it is required to identify the bifurcation and end points of the ridges. These points are referred as minutiae and are also called as feature points. Generation of multi-line code. Once the minutia are extracted from the given fingerprint image, Wong’s technique is applied as shown in Fig. 2a, where one straight line of length (i.e. l) is drawn through the selected minutia. Then, all the sample points of that line are identified and a circle is drawn around it. Now, the minutia coming inside that circle is counted. The total minutia coming inside each circle are arranged in a sequence for generating a line code that describes the reference minutia. The proposed system is illustrated in Fig. 2b, where a single circle will be drawn instead of drawing multiple circles on a line. The total minutia coming under this circle will be counted and used for generating the line code. Permutation in multi-line code. In case of the revocability and template diversity, some external factors must be added to bring variations into the created line code. In case of simple permutation on line code, user defined secret keys are used. The secret key changes the permutation order of the minutia code. The secret keys to be used are generated by using Pseudo-Random Number Generator (PRNG) algorithm. The secret key generated should be unique i.e., it should not be allocated to more than one user/application. This is achieved by using PRNG, which generates unique random numbers.

Privacy Threat Reduction Using Modified Multi-line Code Generation …

283

Fingerprint Matching. The Template Fingerprint (TF) and Query Fingerprint (QF) are used for matching. Every line code from QF is matched with each line code from TF. Based on the result of line code matching, authenticity is checked.

4 Implementation Details The cancelable biometrics are complicated for implementation. The proposed MMLCGA reduces the complexity of multiline code algorithm used in CBT system. In the modified multi-line code, step Line Code Generation converts the fingerprint image into a line, which contains the minutiae in number format. To reduce the complexity of the algorithm, Line Code Generation step is modified. Here, on the line passing through the minutiae, instead of drawing multiples circles over the line only one circle is drawn with large radius. This reduces the computation time of algorithm and improved accuracy of the system. Thus, it helps in reducing the complexity of the existing CBT-MLCG algorithm. The implementation is divided into different steps that are described in short as: • • • • •

Binarization Contextual Filtering Thinning Minutiae Extraction Line Code Algorithm.

4.1 Binarization Image binarization is the process of taking a greyscale image and converting it to black-and-white, essentially reducing the information contained within the image from 256 shades of grey to black and white i.e. binary image. This is sometimes known as image thresholding, although thresholding may produce images with more than 2 levels of grey. It is a form of segmentation, whereby an image is divided into constituent objects. The pseudo code of Binarization is given below: Load the image. Calculate the height and width of image. For loop: Scan the image through its height and width Get the RGB value of point Set lower value to white and higher value to black End loop.

284

P. D. Ganjewar et al.

4.2 Contextual Filtering The contextual filters are mainly used to suppress either the high frequencies in the image i.e., smoothing the image, or the low frequencies i.e., enhancing. A Gaussian filter is a linear filter. It’s usually used to blur the image or to reduce noise. The Gaussian filter will blur edges and reduce contrast.

4.3 Thinning It is commonly used to tidy up the output of edge detectors by reducing all lines to single pixel thickness. Thinning is normally applied to binary images and produces another binary image as output. Here, Zhang-Suen thinning algorithm is used.

4.4 Minutiae Extraction Here, the ridge endings, ridge dot, and bifurcations in image are found marked as minutiae.

4.5 Line Code Algorithm In this step, the image is converted to the line of the numbers like {15, 26, 35, 91, …}. The pseudo code of Line Code Algorithm is given below: Store all the featured points in the list Draw a line through one of the featured points Draw circle on the line with radius r For loop: Scan image for its height and width Find if featured point belongs to the drawn circle If yes, then add that point into Array End loop Permute the calculated line with some random number. So, at the end, the line is generated with the featured points that are present in the drawn circle. The algorithm is implemented in Java, and Fingerprint Verification Competition (FVC) datasets are used for result analysis. Standard FVC datasets are downloaded from http://bias.csr.unibo.it/fvc2004/databases.asp. Here, FVC2000, FVC2002 and FVC2004 datasets are used. Every dataset has fingerprint samples for 10 users. These fingerprint samples are in JPG format.

Privacy Threat Reduction Using Modified Multi-line Code Generation …

285

5 Results and Discussion The existing algorithm (CBT-MLCG) and proposed algorithm (CBT-MMLCG) are implemented on Java platform and tested based on the standard FVC datasets like FVC2000, FVC2002 and FVC2004. This is the extension of work [18]. The user needs to enrol in the system first and will be verified with the existing fingerprint at the time of authentication. The enrollment phase is indirectly a training phase as in, when any new user gets enrolled in the system, training data gets updated and used in the future for authentication. Only the authorized/enrolled users will have the access to the system. The working of the system is shown in Fig. 3a–g. The comparative analysis of existing (CBT-MLCGA) and proposed (CBTMMLCGA) is performed based on the two parameters called as computation time and accuracy of the algorithm. The computation time using CBT-MLCGA and CBTMMLCGA is calculated and shown in Fig. 3e, for one user from FVC2000 dataset. Similarly, it is calculated for all the users from FVC2000, FVC2002 and FVC2004 datasets and recorded in the Table 1 and their comparative analysis is shown in Fig. 4. By looking at the results shown in Table 1, the average time of the proposed technique (CBT-MMLCGA) is approximately 15% less than that of the existing technique (CBT-MLCGA). The accuracy of the algorithm using the existing technique and proposed technique is calculated shown in Fig. 3f, for one user from FVC2000 dataset. Similarly, the accuracies for all the users from FVC2000, FVC2002 and FVC2004 datasets are computed and recorded in the Table 2 and their comparative analysis is shown in Fig. 5. By looking at the results shown in Table 2, the average accuracy of the proposed technique (CBT-MMLCG) is improved by approximately 1.4% over the existing technique (CBT-MLCG).

6 Conclusion Technique like cancelable biometrics has the potential to boost the security and confidentiality of a traditional biometric system. It uses multi-line code algorithm for the verification of the fingerprints. This existing technique is modified, and a new technique called Modified Multiline Code Generation Algorithm, which reduces the complexity of Multiline code algorithm, has been proposed. In the modified multiline code algorithm, the step Line Code Generation convert the fingerprint image into a line, which contains the minutiae in number format. To reduce the complexity of the algorithm, Line Code Generation step is modified by drawing a single circle instead of multiple circles on the line passing through the minutiae. This reduces the computation time of algorithm and improves the accuracy of the system. The algorithm is implemented in Java and analysis have been performed using Fingerprint Verification Competition datasets. Here, FVC2000, FVC2002 and

286

P. D. Ganjewar et al.

Fig. 3 a Initial screen, b training set, c user enrollment, d user authentication, e computation time, f accuracy of authentication using CBT-MLCG, g accuracy of authentication using CBT-MMLCG

Privacy Threat Reduction Using Modified Multi-line Code Generation …

287

Fig. 3 (continued)

Table 1 Time complexity Dataset

FVC2000

FVC2002

User’s

Time complexity (ms)

Time complexity (ms)

FVC2004

CBT-MLCG

CBT-MMLCG

CBT-MLCG

CBT-MMLCG

CBT-MLCG

CBT-MMLCG

User-1

954

779

480

420

1221

1016

User-2

938

851

802

725

839

728

User-3

604

477

493

443

845

714

User-4

276

173

477

417

745

645

User-5

522

460

526

368

673

527

User-6

453

368

493

455

569

487

User-7

451

261

455

383

554

529

User-8

451

374

484

378

549

477

User-9

439

327

462

459

549

476

User-10

188

178

529

443

562

632

Avg. time complexity

527.6

424.8

520.1

449.1

710.6

623.1

Time complexity (ms)

FVC2004 datasets are used. The time and accuracy of the proposed technique (CBTMMLCGA) is better by 15% and 1.4% respectively, when compared to the existing technique (CBT-MLCGA).

P. D. Ganjewar et al.

Average Time (ms)

288

Average Time Complexity 800 600 400 200 0 FVC2000

FVC2002 Database

CBT-MLCGA

FVC2004

CBT-MMLCGA

Fig. 4 Average time complexity Table 2 Accuracy of the algorithms Dataset

FVC2000

User’s

User-1

FVC2002

FVC2004

Percentage accuracy

Percentage accuracy

CBT-MLCG

CBT-MMLCG

CBT-MLCG

CBT-MMLCG

Percentage accuracy CBT-MLCG

CBT-MMLCG

78

80

93

94

64

66

User-2

40

42

55

57

72

73

User-3

56

59

76

79

80

80

User-4

92

93

56

57

93

96

User-5

54

55

94

96

56

57

User-6

52

53

70

71

80

83

User-7

94

95

76

78

82

84

User-8

46

47

76

77

94

95

User-9

65

65

65

67

72

74

User-10

85

84

88

90

92

93

Avg. % accuracy

66.2

67.3

74.9

76.6

78.5

80.1

Percentage

Average Percentage Accuracy 100 80 60 40 20 0 FVC2000

FVC2002 Databases

CBT-MLCG Fig. 5 Accuracy of the algorithms

CBT-MMLCG

FVC2004

Privacy Threat Reduction Using Modified Multi-line Code Generation …

289

Acknowledgements I would like to express my special thanks to my supervisor Dr. Sanjeev. J. Wagh for their constant support and encouragement in my research work. In this work my PG student Ms. Arti Gilbile, helped me for implementation of this algorithm, without her support this would not be possible.

References 1. Kaur H, Khanna P (2019) Random distance method for generating unimodal and multimodal cancelable biometric features. IEEE Trans Inf Forensics Secur 14 2. Wu S-C, Chen P-T, Lee Swindlehurst A, Hung P-L (2019) Cancelable biometric recognition with ECGs: subspace-based approaches. IEEE Trans Inf Forensics Secur 14 3. Jin Z, Lai Y-L, Hwang JY, Kim S, Teoh ABJ (2018) Ranking based locality sensitive hashing enabled cancelable biometrics: index-of-max hashing. IEEE Trans Inf Forensics Secur 13 4. Gomez-Barrero M, Galbally J, Rathgeb C, Busc C (2018) General framework to evaluate unlinkability in biometric template protection system. IEEE Trans Inf Forensics Secur 13 5. Jin Z et al. (2016) Generating fixed-length representation from minutiae using kernel methods for fingerprint authentication. IEEE Trans Syst Man Cybern Syst 46(10):1415–1428 6. Lee C et al. (2007) Alignment-free cancelable fingerprint templates based on local minutiae information. IEEE Trans Syst Man Cybern Part B 37(4):980–992 7. Kelkboom EJC, Molina GG, Breebaart J, Veldhuis RNJ, Kevenaar TAM, Jonker W. Binary biometrics: an analytic framework to estimate the performance curves under Gaussian assumption. IEEE Trans Syst Man Cybern Part A Syst Humans 40(3) 8. Wong WJ, Wong MLD, Kho YH (2013) Multiline code: a low complexity revocable fingerprint template for cancellable biometrics. J Central South Univ 20(5):1292–1297 9. Jin Z, Bok-Min G, Teoh ABJ, Tay YH (2014) Non-invertible analysis on graph-based hamming embedding transform for protecting fingerprint minutiae. In: IEEE 2014 international conference on electronics, information and communications (ICEIC), pp 1–2 10. Jin Z, Lim MH, Teoh ABJ, Goi BM (2014) A non-invertible randomized graph-based hamming embedding for generating cancellable fingerprint template. Pattern Recogn Lett 42:137–147 11. Li M, Feng Q, Zhao J, Yang M, Kang L, Wu L (2014) Minutiae matching with privacy protection based on the combination of garbled circuit and homomorphic encryption. Sci World J. https:// doi.org/10.1155/2014/525387 12. Akdo˘gan D, Altop DK, Levi A (2015) Secure key agreement using pure biometrics. In: 2015 IEEE conference on communications and network security (CNS), pp 191–199 13. Li C et al (2015) A new biocryptosystem-oriented security analysis framework and implementation of multibiometric cryptosystems based on decision level fusion. IEEE Trans Inf Forensics Secur 10(6):1193–1206 14. Subramaniam U, Subbaraya K (2015) A biometric based secure session key agreement using modified elliptic curve cryptography. Int Arab J Inf Technol (IAJIT) 12(2) 15. Barman S, Samanta D, Chattopadhyay S (2015) Fingerprint-based crypto-biometric system for network security. EURASIP J Inform Sec 2015(1):1–17 16. Vijayakumar T (2021) Synthesis of palm print in feature fusion techniques for multimodal biometric recognition system online signature. J Innov Image Process (JIIP) 3(02):131–143 17. Smys S, Wang H (2021) Data elimination on repetition using a blockchain based cyber threat intelligence. IRO J Sustain Wirel Syst 2(4):149–154 18. Gilbile A, Ganjewar PD. Design of secure biometric using cancelable techniques. In: Springer international conference on data science, machine learning and applications (ICDSMLA 2020). EBook ISBN 978-981-16-3690-5, Hardcover ISBN 978-981-16-3689-9. http://doi.org/ 10.1007/978-981-16-3690-5

Systematic Literature Review—IoT-Based Supply Chain Management in Industry 4.0 Sreeparnesh Sharma Sivadevuni and Sathish Kumar Ravichandran

Abstract The twenty-first century has seen considerable implications towards social trends, as well as advances in technology and industrial achievements. The fourth industrial revolution is the consequence of a move towards automation and a decrease in the amount of human participation in most industries (Industry 4.0). Through an in-depth assessment of the existing available work, this article investigates the influence of technology on the Internet of Things (IoT) has and continues to have on Supply Chain Management (SCM) during the era of Industry 4.0. This analysis of the relevant literature not only provides a plethora of fresh information on Industry 4.0 but also indicates areas that require more attention and gives suggestions for the future. Investigating academic developments related to Industry 4.0 will assist in the process of closing this knowledge gap. A comprehensive review of academic articles on the impact of IoT on supply chain management in Industry 4.0 was conducted up to the end of April 2022. Keywords Internet of Things (IoT) · Industrial revolution 4.0 (IR 4.0) · Supply chain management · Systematic literature review (SLR) · State-of-the-art · Qualitative research · Quantitative research · Research agenda

1 Introduction The rapid pace at which economic conditions are changing in the contemporary world is the driving force behind the significance of this research. Automation, Robotics, Digital Technology, Information Technology, Industrial Automation, Augmented Reality, Virtual Reality (AR/VR), and other similar developments are all in response to growing consumer demand and advances in technology. These characteristics are referred to as the “Digital Economy” or “Industry 4.0,” [1]. As a result of increased consumer expectations about the timeliness and level of quality of the services they S. S. Sivadevuni (B) · S. Kumar Ravichandran Department of Computer Science and Engineering, School of Engineering and Technology, Christ University, Bangalore, Karnataka 560001, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_22

291

292

S. S. Sivadevuni and S. Kumar Ravichandran

get, Supply Chain Management (SCM) must now contend with a whole new set of issues [2]. The use of cutting-edge technology is necessary for the sustainable development of the economy over the long term. The acceleration of industrialization throughout the globe is being driven by advances in science and technology around the world [3]. This is particularly true in the industrial sector. There are four main phases of technological evolution that are widely accepted. To date, no one can agree on exactly what constitutes an industrial revolution, but these stages are well-known, nevertheless [4]. It took two centuries for water and steam-powered mechanical manufacturing facilities to give way to electrical mass production, and it took another two centuries for electronic gadgets and information technology to completely automate (IT) [5]. New industrial facilities that were powered by water or steam were the driving force behind each of these revolutions. The Internet of Things (IoT) and Cyber-Physical Systems (CPS) have been noticed by governments and organizations all over the world, and these entities are now making use of the opportunities that these two technological advancements provide [6]. For any sector to be successful in a complex environment, it must have a very agile workforce. The workforce also needs to have a high degree of tolerance and abilities in risk reduction. In addition they must have structural flexibility that enables the persons involved to prepare for rapid reaction to the challenges that they may face with the capability of an organization’s supply chain to adapt to significant changes in the operational environment [7]. However, additional resources, such as buffer inventory and surplus capacity, as well as a higher cost of coordination, are required for greater flexibility and resilience [8]. If companies are going to achieve an appropriate balance between the needed degree of resiliency and flexibility in addition to the associated costs, they need to be able to predict and track the whole supply chain and they must have the necessary velocity to react swiftly to changes. Firms must have a complete awareness of the supply chain to achieve this delicate equilibrium. So-called concepts that can lead supply chain managers are the “4R’s,” which stand for responsiveness, reliability, resilience, and relationships [9, 10].

2 IoT Based Supply Chain Management in Industry 4.0 Supply Chain Management (SCM) is undergoing a paradigm shift because of the Internet of Things (IoT), one of the most recent developments in information technology [11]. Communications within the supply chain are raised to a new level thanks to the Internet of Things, which introduces new possibilities such as human-to-thing communication and autonomous coordination among “things” while they are being stored in a facility or transported between supply chain entities [12]. These new abilities provide a tremendous potential to deal with SCM issues in a more effective way. As a result of the IoT, the supply chain can now be seen with a degree of transparency, agility, and adaptability that has never been possible before [13]. It is possible to get unparalleled insight into all parts of the supply chain when the data released by smart things are successfully gathered, processed, and transformed into

Systematic Literature Review—IoT-Based Supply Chain Management …

293

meaningful information. It is possible that the organization may be made aware of potentially hazardous situations both inside and outdoors due to this increased visibility [14]. New levels of supply chain efficiency may be achieved when these signals are reacted to in a timely way. In the absence of information, it is not the availability of information that has been lacking; rather, it is the technology for collecting and analyzing massive volumes of data and the time lag between the collection of data and action that has been lacking [15]. As a result of the Internet of Things, supply chains will be able to react to changes in real-time and reach historically unachievable levels of agility and adaptability, allowing them to adapt to market demands in real-time [16]. Remote management of supply chain operations, better cooperation with partners, and the provision of more exact information to assist in more effective decision-making will all be made possible thanks to the Internet of Things [17]. After conducting a brief survey of literature, this article discusses the IoT and the influence it has on SCM in the age of Industry 4.0. This study covers major features of IoT in SCM including the concept of IoT, the primary IoT technological pieces required in its implementation in a supply-chain context, numerous SCM applications, aspects of contemporary manufacturing, and applications that are currently applied [18]. To provide an answer which is more suitable to the stated main research question, the purpose of this work is to review and analyze the academic progress in a methodical manner. This will allow us to provide an answer that is more relevant to the stated main research question. Because of this, we will be able to deliver an answer that is more relevant to the primary research topic that was posed. The first stage result from a larger research effort that will include reviews of other high-level projects and plans related to the fourth industrial revolution around the world [3]. These reviews will be included in the larger research effort. As a component of this more extensive research initiative, this review will be given.

3 Principles and Methods 3.1 Review Principles Three essential review guidelines were created to guarantee that all articles could be appraised uniformly with fewer subjective opinions: Inclusion and exclusion criteria must be clearly stated [19]. Each manuscript should have a clear set of inclusion/exclusion criteria. The following segregation shows six key inclusion and exclusion criteria, as well as their subgroups, in this review. Exclusions 1. The full text of the document is not accessible in English; just the paper’s title, its abstract, and its important phrases are translated into the language. 2. An evaluation of a paper that is incomplete.

294

S. S. Sivadevuni and S. Kumar Ravichandran

3. Articles that are not scholarly in nature do not belong in this category. Examples include editorials, conference assessments, the table of contents, and prefaces. 4. IoT, SCM and Industry 4.0 review, survey, debate, and issue solution are not the subject of a paper. IoT, SCM and Industry 4.0 is merely given as an example to illustrate the point. It is only in the context of a future research direction, viewpoint, or need that the terms “IoT, SCM and Industry 4.0” is employed [6]. A reference to IoT, SCM and Industry 4.0 can only be found in a few places. It is solely used as a phrase or reference in relation to IoT, SCM, and Industry 4.0. Inclusions 5. Without IoT, SCM and Industry 4.0, a study of the fourth industrial revolution’s role in SCM will be incomplete. If you’re writing an academic article is written for utilization of IoT, SCM and Industry 4.0 as a way of describing some of the difficulties, concerns, or trends are necessary to cover its importance. 6. A clear and unambiguous focus on IoT, SCM and Industry 4.0 may be seen throughout the research activities presented in this paper.

3.2 Systematic Literature Review Method It was necessary to conduct an extensive study of Industry 4.0 in IoT and SCM owing to this issue, as well as the results of previous studies [20]. The review for this article’s goals was conducted using a structured review technique. Methods for systematic reviews typically include five stages: formulation of research questions, Locating study, selection and evaluation of study sites, analysis and synthesis of results, and reporting and implementation of findings [21]. These five stages guided the selection of this approach (Fig. 1). Step 1: Question Formulation As a first step, research on IoT and associated difficulties in supply chain logistics began with a study of the number of studies on the topic as well as a review of their context and methodology [22]. They also looked at previous studies to see what had worked and what had not to come up with a list of suggestions. For the purposes of data collection and analysis, two questions are presented in Fig. 2. Quantity of Evaluated and Chosen Studies in Each Database is shown in Table 1. Step 2: Locating Studies Finding, selecting, and evaluating relevant research to answer the review’s specific questions was accomplished. Google Scholar was utilized as the initial search engine since it displays the most results from all databases. The five keyword phrases used were “Industry 4.0 and Supply Chain,” “Industry 4.0 and IoT,” “IoT in Supply Chain,” “Smart Supply Chain,” and “E-Supply Chain” [23]. The title, abstract, or keywords included in the publications were evaluated to locate related studies. As the last step, it was decided to do an extensive search of the main research databases, such as

Systematic Literature Review—IoT-Based Supply Chain Management …

295

Formula on of Research ques on

Synthesis of results and repor ng

Loca ng of study

Analysis and systhesis of results

Selec on and evalua on of study sites

Fig. 1 The systematic review in current research is outlined below Question 1:

Industry 4.0's IoT supply chain trends: what can we expect?

Critiria Analysis: methods

How many studies have been done? Publication dates a repository for research

Question 2:

Will the Internet of Things make it easier to plan, control, and coordinate the operations involved in supply chain management?

Critiria Analysis:

These categories were chosen based on a combination of factors including consensus among experts, methodological approaches and opportunities for additional study.

Fig. 2 Question formation and criteria analysis Table 1 Quantity of evaluated and chosen studies in each database

Data base reviewed Taylor and Francis

Number of papers selected 9

IEEE

16

Springer

14

Emerald

6

Wiley

8

MDPI

4

Total

57

296

S. S. Sivadevuni and S. Kumar Ravichandran

Taylor and Francis, Emerald, Elsevier, IEEE, MDPI and Springer [24]. A lack of relevant articles could not be found due to restrictions imposed by keyword selection and database listings. All the review articles in this study were published after 2014. since the subject matter is still relatively fresh. To make matters worse, only articles published in English were considered for inclusion. Step 3: Study Selection and Evaluation The researchers analyzed the text of each manuscript to determine which research was relevant to this subject. They chose articles that cover the supply chain in Industry 4.0-addressed smart factories [25]. Wiley and Semantic Scholar databases were used since there were not enough publications published in this field and further study was required. This study was not found during the search of the chosen papers, but they picked relevant published articles that had possible material related to this research. A total of 57 of the 507 articles were chosen for the study. Step 4: Analyzing and Synthesizing Using a set of pre-formulated questions, each individual research was discussed. When looking for trends in existing research on Industry 4.0 and supply chain, we looked for things like how many articles have been published since a certain date and where in the world they were conducted [8]. We also looked for things like research methods like surveying and interviewing as well as how much content was investigated and what previous research had been done on the topic was also taken into account. An analysis of gathered articles was performed to classify them as exploratory, confirmatory, qualitative, or process/technology-level research to provide a response to the second research question. There are many types of papers: exploratory, quantitative, and process-level [6]. In addition, the non-negative matrix factorization (NMF) approach was employed to construct words that correctly characterize each cluster using TM. Human expert and machine-based methodologies are used to provide the following findings and analyses [26]. Step 5: Reporting and Using Results Researchers provide their findings based on a technique that includes the assessment of chosen articles according to categories described in the previous paragraph and TM. Based on this evaluation, research gaps are identified and suggestions for future studies are offered for improvement. A summary and conclusions are provided at the end of this publication [27].

4 Review of Research Trends Figure 3 displays the publication schedule from 2016 to 2022, as shown. IoT-based Industry 4.0 supply chain research began in 2016, and the number of studies on this topic is predicted to grow dramatically by the end of 2023. Since just a few articles were approved for publication in July 2022, only half of this year’s data reflects this

No. of publications

Systematic Literature Review—IoT-Based Supply Chain Management …

297

15 10 5 0 2016

2017

2018

2019

2020

2021

2022

Year

No. Of publications

Fig. 3 Number of publications distributed by year

30 25 20 15 10 5 0 Interview

Case study

Content Analysis

Literature Review

Modeling

Research Method Applied Fig. 4 Research method applied

year’s trend. With a total of 8 papers published, IEEE had the most publications (see Fig. 3) followed by Springer with 9 publications. This demonstrates that the main publishers and repositories have prioritized and covered Industry 4.0 in the supply chain. Prior research has mostly used content analysis to study Industry 4.0 with IoT in the SC (Fig. 4), indicating that other empirical approaches like surveys, interviews, case studies, modelling, etc. are hardly used.

5 Content Analysis Two distinct approaches were used to perform a content analysis to address the study’s second goal, which was to identify research-relevant topics and areas. The first method relied on human expert analysis; the second method used traditional machine learning (TM) [28] and to find a solution to the question, we applied both methods. In addition to this, the most pressing problems, current trends, and recent findings are emphasized. Is there a distinction between the conceptualization of Industry 4.0 and its actual implementation in the supply chain? According to the findings of the investigation, there is no clear solution [29]. The authors were not successful in their search for a complete literature classification for the supply chain that was based on IoT based SCM in Industry 4.0. As a direct consequence of this, the authors did a comprehensive review of several different research. The several types of studies, including exploratory, confirmatory, qualitative, and quantitative, as well as management, process, and technology-level research, were each broken

298

S. S. Sivadevuni and S. Kumar Ravichandran

down further into three categories [30]. Most of the articles that were selected for this investigation are review papers. These papers are of a format that is comparable to the current investigation; however, they focus on different topics, use different methods, and cover the different subject matter, which disqualifies them from consideration in the management versus technology level categories. Because of this, it was decided that they would not be considered in this part of their investigation [31]. Ben Daya, for instance, examined both the significance of the Internet of Things to the SC as well as its position in the organization [32]. In one of the studies, a framework was developed that might be used to detect I4 in the construction supply chain (CSC). Zhong et al. published another piece of study on Intelligent Manufacturing in the age of 14, which can be found here [33]. The research carried out by Barreto and colleagues investigates the effects that has on the logistics industry [34]. Figure 5 presents the analysis level, which includes a summary of the classified papers that fall under the management and technology level and is based on both qualitative and quantitative analysis. Fig. 5 Analysis level

30 25 20 15 10 5 0 Management level Qualitative

Process\Technology level Quantitative

30 25 20 15 10 5 0 Management Level

Exploratory

Process\Technology level Confirmatory

Systematic Literature Review—IoT-Based Supply Chain Management …

299

Management Level Several publications at the “management level” pushed for an industrialization and integration framework for the supply chain. Industrialization and supply chains were explored in this research. Researchers predict that the traditional supply chain will be rendered obsolete by emerging technologies such as cyber-physical systems, the Internet of Things (IoT), the iPhone/iPad/iPod Touch, smart manufacturing, and big data. It has been determined that Industry 4.0 is a platform for the exploration of new digital technologies, process optimization, and digitalization. Process/Technology Level Industry 4.0 is said to be based on dynamic systems that are driven by real-time data. Some of the research focused on the dynamic SC. Researchers led by Ivanov devised a mechanism for automating manufacturing facilities’ intelligent supply chain planning schedules. Supply Networks of the Future’s dynamic service scheduling was conceptualized by Vladimirovich Sokolov and his colleagues in 2017 [35]. Research undertaken by Dunke et al. examined how online optimization can be used to address real-time difficulties in Industry 4.0 and SC planning [31].

6 Industry 4.0 and IoT in Supply Chain Research (Future Work) In the current investigation, the content analysis of the selected publications is discussed in two different ways: first, based on the authors’ systematic review, and second, based on the TM that was used to discover research gaps and future research prospects [36]. Both methods are described in detail below. (a) Most of the technical publications in this review highlight a conceptual or technological framework. More technical papers explaining the approach and technology implementations are needed [37]. (b) Few studies highlight Industry 4.0’s influence on the supply chain and evaluate it using before-and-after case studies. It’s also important to know what effects Industry 4.0 will have on firm productivity [38]. (c) Few quantitative studies exist, indicating a lack of analytical research and technological know-how [39]. (d) In addition to the three categories (exploratory, confirmatory, and qualitative/quantitative), each topic may be sorted into one of the following three categories: management, process, and technology [40].

300

S. S. Sivadevuni and S. Kumar Ravichandran

7 Conclusion Using a thorough literature analysis, the authors of this work assessed the current state of research on the use of Industry 4.0 in IoT based supply chain management. The results of this study’s literature assessment demonstrate that the SC views Industry 4.0 as an important idea [41]. Using this notion, human contact would be reduced, and efficiency would be boosted in enterprises [42]. Engineers and academics may use these sectors as a starting point for Industry 4.0 deployment and study in these areas. A few of them are worthy of additional study and examination [43]. Industry 4.0 experts are few and far between, which hampers further investigation and conversation on the issue [44].

References 1. Abdirad M, Krishnan K (2020) Industry 4.0 in logistics and supply chain management: a systematic literature review. EMJ Eng Manag J 1–15. http://doi.org/10.1080/10429247.2020. 1783935 2. Full article: 5G in digital supply chain and operations management: fostering flexibility, endto-end connectivity and real-time visibility through internet-of-everything. https://doi.org/10. 1080/00207543.2021.2002969. Accessed 31 May 2022 3. Machado CG, Winroth MP, Ribeiro da Silva EHD (2020) Sustainable manufacturing in Industry 4.0: an emerging research agenda. Int J Prod Res 58(5):1462–1484. https://doi.org/10.1080/ 00207543.2019.1652777 4. (PDF) The Industry 4.0 revolution and the future of manufacturing execution systems (MES). https://www.researchgate.net/publication/306150248_The_Industry_40_rev olution_and_the_future_of_Manufacturing_Execution_Systems_MES. Accessed 08 June 2022 5. Mourtzis D, Vlachou E, Zogopoulos V, Fotini X (2022) Advances in production management systems. The path to intelligent, collaborative and sustainable manufacturing, vol 513, pp 354–362. Accessed: 08 June 2022 [Online]. Available: https://doi.org/10.1007/978-3-319-669 23-6 6. Fatorachian H, Kazemi H (2021) Impact of Industry 4.0 on supply chain performance. Prod Plan Control 32(1):63–81. https://doi.org/10.1080/09537287.2020.1712487 7. Wamba SF, Queiroz MM (2022) Industry 4.0 and the supply chain digitalisation: a blockchain diffusion perspective. Prod Plan Control 33(2–3):193–210. https://doi.org/10.1080/09537287. 2020.1810756 8. Ivanov D, Dolgui A, Sokolov B (2019) The impact of digital technology and Industry 4.0 on the ripple effect and supply chain risk analytics. Int J Prod Res 57(3):829–846. https://doi.org/ 10.1080/00207543.2018.1488086 9. Zheng T, Ardolino M, Bacchetti A, Perona M (2021) The applications of Industry 4.0 technologies in manufacturing context: a systematic literature review. Int J Prod Res 59(6):1922–1954. http://doi.org/10.1080/00207543.2020.1824085 10. Álvarez-Rodríguez DA, Normey-Rico JE, Flesch RCC (2017) Model predictive control for inventory management in biomass manufacturing supply chains. Int J Prod Res 55(12):3596– 3608. https://doi.org/10.1080/00207543.2017.1315191 11. Mishra A, Jha AV, Appasani B, Ray AK, Gupta DK, Ghazali AN (2022) Emerging technologies and design aspects of next generation cyber physical system with a smart city application perspective. Int J Syst Assur Eng Manag. http://doi.org/10.1007/s13198-021-01523-y

Systematic Literature Review—IoT-Based Supply Chain Management …

301

12. Xu H, Yu W, Griffith D, Golmie N (2018) A survey on industrial internet of things: a cyberphysical systems perspective. IEEE Access 6:78238–78259. http://doi.org/10.1109/ACCESS. 2018.2884906 13. Da Xu L, Xu EL, Li L (2018) Industry 4.0: state of the art and future trends. Int J Prod Res 56(8):2941–2962. https://doi.org/10.1080/00207543.2018.1444806 14. Yang H, Kumara S, Bukkapatnam STS, Tsung F (2019) The internet of things for smart manufacturing: a review. IISE Trans 51(11):1190–1216. https://doi.org/10.1080/24725854.2018.155 5383 15. Sethi P, Sarangi SR (2017) Internet of things: architectures, protocols, and applications. J Electr Comput Eng 2017. http://doi.org/10.1155/2017/9324035 16. Ghasempour A (2019) Internet of things in smart grid: architecture, applications, services, key technologies, and challenges. Inventions 4(1). http://doi.org/10.3390/inventions4010022 17. Wang R, Wang D, Wang L, Dou Y (2021) Design of remote monitoring system for new energy vehicles from the perspective of key technologies. In: 2021 IEEE international conference on advances in electrical engineering and computer applications, AEECA 2021, Aug 2021, pp 463–467. http://doi.org/10.1109/AEECA52519.2021.9574236 18. Petrenko AS, Petrenko SA, Makoveichuk KA, Chetyrbok PV (2018) The IIoT/IoT device control model based on narrow-band IoT (NB-IoT). In: Proceedings of the 2018 IEEE conference of Russian young researchers in electrical and electronic engineering, ElConRus 2018, vol 2018, pp 950–953. http://doi.org/10.1109/EIConRus.2018.8317246 19. Li S, Da Xu L, Zhao S (2018) 5G internet of things: a survey. J Ind Inf Integr 10:1–9. http:// doi.org/10.1016/j.jii.2018.01.005 20. Hassan MB, Ali ES, Mokhtar RA, Saeed RA, Chaudhari BS (2020) NB-IoT: concepts, applications, and deployment challenges. In: LPWAN technologies for IoT and M2M applications. Elsevier, pp 119–144. http://doi.org/10.1016/b978-0-12-818880-4.00006-5 21. Elgarhy O, Reggiani L (2018) Increasing efficiency of resource allocation for D2D communication in NB-IoT context. Procedia Comput Sci 130:1084–1089. https://doi.org/10.1016/j. procs.2018.04.160 22. Kusiak A (2020) Open manufacturing: a design-for-resilience approach. Int J Prod Res 58(15):4647–4658. https://doi.org/10.1080/00207543.2020.1770894 23. Buer SV, Strandhagen JO, Chan FTS (2018) The link between industry 4.0 and lean manufacturing: mapping current research and establishing a research agenda. Int J Prod Res 56(8):2924–2940. https://doi.org/10.1080/00207543.2018.1442945 24. Raza U, Kulkarni P, Sooriyabandara M (2017) Low power wide area networks: an overview. IEEE Commun Surv Tutorials 19(2):855–873. https://doi.org/10.1109/COMST.2017.2652320 25. Dolgui A, Ivanov D (2022) 5G in digital supply chain and operations management: fostering flexibility, end-to-end connectivity and real-time visibility through internet-of-everything. Int J Prod Res 60(2):442–451. https://doi.org/10.1080/00207543.2021.2002969 26. Dolgui A, Ivanov D, Sokolov B (2020) Reconfigurable supply chain: the X-network. Int J Prod Res 58(13):4138–4163. http://doi.org/10.1080/00207543.2020.1774679 27. Moghaddam M, Nof SY (2018) Collaborative service-component integration in cloud manufacturing. Int J Prod Res 56(1–2):677–691. https://doi.org/10.1080/00207543.2017.137 4574 28. Ivanov D (2021) Supply chain viability and the COVID-19 pandemic: a conceptual and formal generalisation of four major adaptation strategies. Int J Prod Res 59(12):3535–3552. https:// doi.org/10.1080/00207543.2021.1890852 29. Sasi A, Subramanian T, Kumar Ravichandran S (2022) Systematic literature review on industry revolution 4.0 to enhance supply chain operation performance, pp 173–179. http://doi.org/10. 1145/3512676.3512705 30. Javidroozi V, Shah H, Feldman G (2019) Urban computing and smart cities: towards changing city processes by applying enterprise systems integration practices. IEEE Access 7:108023– 108034. https://doi.org/10.1109/ACCESS.2019.2933045 31. Brintrup A et al (2020) Supply chain data analytics for predicting supplier disruptions: a case study in complex asset manufacturing. Int J Prod Res 58(11):3330–3341. https://doi.org/10. 1080/00207543.2019.1685705

302

S. S. Sivadevuni and S. Kumar Ravichandran

32. Ben-Daya M, Hassini E, Bahroun Z (2019) Internet of things and supply chain management: a literature review. Int J Prod Res 57(15–16):4719–4742. http://doi.org/10.1080/00207543.2017. 1402140 33. Zheng P et al (2018) Smart manufacturing systems for Industry 4.0: conceptual framework, scenarios, and future perspectives. Front Mech Eng 13(2):137–150. http://doi.org/10.1007/s11 465-018-0499-5 34. Barreto L, Amaral A, Pereira T (2017) Industry 4.0 implications in logistics: an overview. Procedia Manuf 13:1245–1252. https://doi.org/10.1016/j.promfg.2017.09.045 35. Ivanov D, Sokolov B, Chen W, Dolgui A, Werner F, Potryasaev S (2021) A control approach to scheduling flexibly configurable jobs with dynamic structural-logical constraints. IISE Trans 53(1):21–38. https://doi.org/10.1080/24725854.2020.1739787 36. Rao SK, Prasad R (2018) Impact of 5G technologies on Industry 4.0. Wirel Pers Commun 100(1):145–159. https://doi.org/10.1007/s11277-018-5615-7 37. Ivanov D, Dolgui A (2021) Stress testing supply chains and creating viable ecosystems. Oper Manag Res. https://doi.org/10.1007/s12063-021-00194-z 38. Dolgui A, Ivanov D, Potryasaev S, Sokolov B, Ivanova M, Werner F (2020) Blockchain-oriented dynamic modelling of smart contract design and execution in the supply chain. Int J Prod Res 58(7):2184–2199. https://doi.org/10.1080/00207543.2019.1627439 39. Jouini W, Moy C, Palicot J (2012) Decision making for cognitive radio equipment: analysis of the first 10 years of exploration. Eurasip J Wirel Commun Netw 2012. http://doi.org/10.1186/ 1687-1499-2012-26 40. Wu D, Rosen DW, Wang L, Schaefer D (2015) Cloud-based design and manufacturing: a new paradigm in digital manufacturing and design innovation. CAD Comput Aided Des 59:1–14. https://doi.org/10.1016/j.cad.2014.07.006 41. Zhong RY, Xu C, Chen C, Huang GQ (2017) Big data analytics for physical internet-based intelligent manufacturing shop floors. Int J Prod Res 55(9):2610–2621. https://doi.org/10.1080/ 00207543.2015.1086037 42. Kumar Y, Singla R (2022) Effectiveness of machine and deep learning in IOT-enabled devices for healthcare system. In: Internet of things, pp 1–19. http://doi.org/10.1007/978-3-030-814 73-1_1 43. Ivanov D (2020) Viable supply chain model: integrating agility, resilience and sustainability perspectives—lessons from and thinking beyond the COVID-19 pandemic. Ann Oper Res. https://doi.org/10.1007/s10479-020-03640-6 44. Chen M, Miao Y, Hao Y, Hwang K (2017) Narrow band internet of things. IEEE Access 5:20557–20577. https://doi.org/10.1109/ACCESS.2017.2751586

A Review on Urban Flood Management Techniques for the Smart City and Future Research Anil Mahadeo Hingmire and Pawan R. Bhaladhare

Abstract Flooding in cities is a worldwide occurrence that presents a significant problem to municipal administrations and urban planners. The loss of the life, delays in public transportation, damage to public and private property, the interruption of services such as the water supply and power supply are some of the effects of urban flooding which leads to economic losses as well as public health issues. The motive of this research paper is to review the various strategies for managing urban floods and to determine the research scope in terms of smart city development. The flood is one of the most prevalent natural catastrophes that may strike any city. Rainfall, water level, temperature, humidity, drainage water level, water discharge, as well as other parameters are generally viewed in flood prediction models including artificial neural networks (ANN), fuzzy inference processes, regression models, deep learning, gradient boosting decision trees, and self-organizing feature mapping networks (SOM). Real-time flood parameters were considered in the flood detection and warning system. Real-time flood characteristics were considered in the flood detection and warning system, and the system was constructed utilizing IoT. The accuracy of flood prediction of computational intelligence techniques is only 76.48% in average. Keywords Urban flood · Smart city · Flood forecasting · Flood detection · Flood warning · Artificial neural network · K-nearest neighbor · Machine learning · Internet of technology

1 Introduction Millions of people across the globe are afflicted by flood disasters, which result in substantial loss of life and property devastation. Urban flooding is becoming more common across the world, posing a serious challenge to city governments and urban A. M. Hingmire (B) · P. R. Bhaladhare School of Engineering, Sandip University, Nashik, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_23

303

304

A. M. Hingmire and P. R. Bhaladhare

planners. Urban floods cause difficulties ranging from small occurrences to major occurrences, with cities being inundated for a few hours to many days. In developing nations, urbanization increased from less than 25% in 1970 to more than 50% in 2006. By 2020, seven of the top 10 economies in the world will be located in Asia. At the same time, Asia is one of the world’s most rapidly urbanizing areas. Peak flows accelerate the occurrence of floods, which affect large populations in densely populated areas and result in catastrophic infrastructural and economic losses for business and industry. The climate model of Rafiq et al., predicted that the rainfall in winter would blow-up to the 20–30% by the 2080s. There are two types of flood causes: direct and indirect. Natural reasons such as global climate change, changes in weather patterns, high rainfall, and so on, as well as man-made ones such as the loss of natural drains and continual urban development owing to population growth, are examples of direct causes. Indirect factors include an insufficient and faulty drainage system, on-street parking, and a poor or non-existent waste management system. Global climate change, urbanization, uncontrolled dam water releases, and inefficient and insufficient drainage systems are among the leading causes of floods, according to the report. The global climate is made up of the sun, earth, seas, wind, rain, snow, forests, deserts, and other interrelated systems. Due to fluctuating weather patterns, we can’t predict rain, snow, storms, cloudbursts, and other weather occurrences. Glaciers are receding on average at a rate of 10–15 m per year, according to a National Intelligence Council report and international studies of glaciers melting due to climate change. If the rate accelerates, river basins fed by these glaciers may flood, resulting in lower flows and water shortages for drinking and cultivation. The amount of impervious surface area in a given site increases due to urbanization, resulting in decreasing of hydrologic response time and the risk of flood. Much of the sewage and drainage system’s condition is unknown. They are either overwhelmed by the water or choked with garbage and non-biodegradable plastic bags. Because of unlawful connections, sewers overflow, and the sewage system can no longer handle the increasing volume. The sudden and unexpected release of water from dams and lakes, which occurs without giving the populace the time to react, causes floods in metropolitan areas. Because most urban drainage systems were built to handle lesser amounts of runoff than those encountered today, they are prone to failure during high-intensity rainstorm storms. During a period of severe weather, even the world’s brightest cities will be put to the test by floods. As a result, smart city administrations will need to take a comprehensive approach to preventing such tragedies. Figure 1 depicts the flood management process, which includes real-time risk management, pre-flood measures, and post-flood measures. Pre-flood precautions include things like insurance, flood defense measures, and spatial planning. Risk management in real time encompasses forecasting, warning, and rescue, among other things. Including relief, cleanup, reconstruction, regeneration, etc., are post-flood measures. Sensors and AI based algorithms can aid in the development of appropriate levels of monitoring and reaction, whether it’s for calibrating dam water releases, controlling rivers, sewage network etc. Other options include storm surge obstacles, flood

A Review on Urban Flood Management Techniques for the Smart City …

305

Fig. 1 Flood management

insurance, 3D mapping and simulations, enhanced urban design, and regulatory compliance. The aim of this paper is to analyze the key urban flood management techniques that have been proposed and implemented to mitigate urban flooding. The study’s other goal is to determine the scope of future research in the subject of urban flood management. Engineers, designers, and administrators can use this review to get a broad understanding of flood management approaches for smart cities.

2 Related Work Munawar et al. [1] focuses on flood detection using classifier model in deep learning with specific implementation of Harr cascade classifier approach on Unmanned Aerial Vehicles (UAVs) images. The 300 UAV images used to classify the flooded and non-flooded regions from the intended area. The image preprocessing, a landmark feature selection for detection, model training, employing image classification to detect floods and measuring the effectiveness of the recommended model. In these experimental results the detection roads and landmark accuracy for buildings were found 94% and 91% respectively. The accuracy observed is 91% for the given UAV input images which can be used for flood relief and rescue operation. The limitations or research gaps in the manuscript is focusing on detection of 2D flood-affected areas. A region’s flood severity may be determined by measuring the depth of the floodwaters [1]. Darabi et al. [2], the work in this paper focusing on the spatial prediction of urban flood using hybrid multi-boosting neural network model (MultiB-MLPNN) and testing in the flood prone area. The comparative study of MultiB-MLPNN and

306

A. M. Hingmire and P. R. Bhaladhare

Multilayer Perceptron Neural Network (MLPNN) is performed and observed that the performance of the hybrid multi-boosting neural network model (MultiB-MLPNN) is better than the MLPNN [2–4]. The limitation or research gap in this article is that hydraulic data such as inundation depth, velocity, etc. was not available to hydraulicsbased models for the study area. Pham et al. [5] focuses on the Flood risk assessment using a hybrid model, which comprises deep learning and Multi-Criteria Decision Analysis model (MCDA). The model is trained and tested with the 847 data records of past flood locations in the study area. Deep Neural Networks (DNNs) algorithm is used to create a map of flood susceptibility, and the MCDA technique was utilized to create maps of hazards, exposure, and vulnerabilities. The model uses the geospatial databases which include curvature, distance from the river, elevation, flow accumulation, river density, rainfall, land cover and location of the flood for training purpose [5]. The limitation of this model is that it requires local geo environmental dataset for implementations at different locations. Samikwa et al. [6] developed a model for flood prophecy using Artificial Neural Network (ANN), Internet of Things (IoT) and edge computing technology. The system leverages temporal correlative information from water level data as well as the real-time rainfall data to anticipate flood water levels ahead of time utilizing long short term memory [6]. The data collected of the water level and rainfall using the distance and rainfall sensors, respectively. The sensors are immediately linked to an Arduino Nano 33 BLE, a low power IoT BLE device. Researchers used a collection of data from Melbourne Water that included hourly rainfall and water level data. The dataset, which has 78,844 rows and was compiled hourly between 2009-12-01 00:00:00 and 2018-12-02 18:00:00, contains statistics on rainfall and water levels for around 9 years. The research gap or limitation of the model is the intensity of urban flood is not determined in the research [6]. Wu et al. [7] designed a model for the prophecy of urban flood depth based on rainfall returns period dataset using deep learning. Gradient Boosting Decision Tree (GBDT) is used in this work to forecast the depth of an urban flooded region using a regression model [7–9]. The model is based on water evaporation, rainfall duration, peak rainfall, catchment area, slope and land use such as roads, buildings, grassland, etc. The model’s evaluation revealed that the relative prediction error was 11.52%, demonstrating the method’s relevance to the depth prediction of urban floods [7]. The research gap or limitation is it relies on conditional factors such as slope, land use, buildings not on the real time driving factors like heavy rainfall, high soil moisture, water velocity, volume etc. Yoon [10] focuses on urban flood forecasting based on Numerical Weather Prediction using quantitative precipitation forecasts (QPFs) method. In order to increase the precision of the forecasts, this study used the Adaptive Blending Method, which incorporates radar-based extrapolation and QPFs like the McGill Algorithm for Prediction Nowcasting by Lagrangian Extrapolation (MAPLE), the KOrea NOwcasting System (KONOS), the Spatial-scale Decomposition method (SCDM), Unified Model Local Data Assimilation and Prediction System (UM LDAPS), and Advanced Storm-scale Analysis and (ASAPS) [10–14]. This study uses radar-based

A Review on Urban Flood Management Techniques for the Smart City …

307

QPF data that have a geographical resolution of 1 km and a temporal resolution of 10 min. The Korea Meteorological Administration (KMA) releases QPF data for the Seoul metropolitan region, which makes it simple to assess the hydrological applicability and accuracy of different QPFs. The necessity to examine more diversified rainfall cases and study locations, as well as optimize weights utilizing various storm events need to be considered. Yves Abou Rjeily et al. focus on establishing a flooding forecast system (FFS) that can inform the Urban Drainage System authority for probable floods ahead of time in 2017 [15, 16]. For forecasted storm events, the model estimates the depth of water variation in the critical manholes. Proposed FFS and Nonlinear Auto Regressive with eXogenous Inputs (NARX) neural network are used in the strategy to predict flooding episodes. The NARX (Nonlinear Auto Regressive with eXogenous Inputs) neural network calculates its output by fusing exogenous input with recurrent behaviour. The NARX neural network’s exogenous input can represent a time series of rainfall intensity [17–20]. The limitation of this research is that it uses the small size of dataset as it considered five years data and few manholes. Chen et al. [21] proposed a flood inundation index-based urban flood forecasting model which contains a runoff production module that categorizes surfaces as pervious or impervious, as well as a surface runoff routing method that routes surface runoff and calculates the inundation index [21]. This work offers the DPSIR (Driving Pressure Index, Impact Index, Force Index, State Index, and Response Index) urban flood warning approach, which is based on multiple index fuzzy assessment. The urban flood forecasting model includes the precipitation prediction, pipeline runoff routing, runoff production, surface runoff routing, drainage system and grid division. The approach described in the research performs better but does not concentrate on the flood routing model. Simões et al. [22] represented a simulation work of 1D/1D, 1D/2D approach for urban drainage flood risk assessment and its performance. The paper includes the development of hydraulic 1D/1D, 1D/2D and Hybrid Infoworks of overland and sewer networks for drainage flood forecasting and calculating the computational time [23–26]. Because of the extended simulation duration, this study work’s restriction or conclusions are that it is not yet suited for use in real-time flood prediction applications. A new sort of model was given to overcome this flaw. It combines a 1D and a 2D overland network, enabling 2D simulation in the most important locations with a quick simulation time and no information loss in other flooded areas [22]. In 2014, Lohani et al. created a modified fuzzy inference approach for real-time flood forecasting. By incorporating the notion of unusual and common hydrological conditions into the fuzzy modelling system, which consists of a modified Takagi Sugeno (T-S) fuzzy inference system known as the threshold subtractive clustering based Takagi Sugeno (TSC-T-S) fuzzy inference system [27, 28]. To evaluate the effectiveness of a flood forecasting model the peak percent threshold statistics (PPTS) used as a performance criteria. Using hourly rainfall and discharge data, the created model was evaluated for various lead durations [29]. The limitations of this research is that it focuses on flood forecasting and false prediction, not on warning as well as flood routing technology.

308

A. M. Hingmire and P. R. Bhaladhare

Naik et al. [30] implemented and analyzed logistic regression method to predict the flood in Kerala state. In this study, a rainfall dataset of Kerala state for last 115 years was utilized to train and forecast a flood model. The dataset considered is month wise rainfall and is from data.gov.in. The accuracy of the model was found 75% only [30]. The model’s shortcoming is that the flood forecast accuracy is weak, and it has to be improved. Mendoza-Cano et al. developed and tested the water level monitoring and flood alerting system based on IoT-based sensor network in 2021. The project used IoT and a wireless sensor network to monitor and warn about urban flooding in the Mexican city of Colima-Villa de alvarez [31]. During extreme occurrences such as tropical storms, the technique comprises a smart water network that collects real-time hydro meteorological data. For flood monitoring, the approach takes into account the soil moisture, fluvial water level, and meteorological data [32]. From a hydrological standpoint, the interconnections between tributaries and crucial sites for data collecting essential for monitoring and modelling were also taken into account. Each node has the relevant sensors, a data logger, communication module, power module, and local backup store. The system lacks a flood resilience notion while being inexpensive. Chang and Chang [33] proposes Artificial Intelligence based Urban Flood Warning System using IoT-based Flood Depth Sensors. For continually learning and updating model parameters, the technique combines a regional flooding prediction algorithm with an online correction algorithm. This study combines the recurrent nonlinear autoregressive with exogenous inputs network (R-NARX) model with selforganizing feature mapping networks (SOM) to forecast regional floods [33]. The study focuses on flood prediction rather than flood resistance. Vinothini and Jayanthy [34] used the Internet of Things to create a real-time Flood Detection and Notification System. The system uses a sensor network to gather data from the ground field, such as temperature, humidity, and water levels, and a Decision Tree classifier to analyse flood information in order to determine if the amount of water is normal or dangerous [34]. The DHT11 sensor is used for temperature and humidity sensor. The probe sensor is used to measure water level with PIC16F877 microcontroller. The raw data is collected from sensor values is 120 flood values. The research reveals that the suggested Decision Tree Algorithm outperforms the existing HyperPipes Algorithm in terms of accuracy. The suggested system’s weakness is that it was only tested on a short dataset (only 120 flood values), and it has to be evaluated on a bigger dataset. By comparing the data from anticipated and actual flood occurrences, it is possible to assess the accuracy of the flood forecasting model [35–37].

3 Analysis and Discussion Analysis of urban flood management techniques has been performed in this section. Table 1 show the analysis of the techniques.

Author(s)

Hafiz et al. [1]

Darabi et al. [2]

Pham et al. [5]

Samikwa et al. [6]

S. No.

1

2

3

4

UAV dataset, deep learning, disaster management, landmarks detection

Key points in the coverage Haar cascade classifier, CNN classifier

Techniques used

Flood prediction, early flood warning

Flood risk assessment

Edge computing, flood prediction, artificial neural networks, Internet of things, long short-term memory

Multi-criteria decision analysis, food risk assessment, AHP, deep learning

Artificial Neural Network (ANN), short-term flood prediction

Deep Neural Networks (DNNs), Multi-Criteria Decision Analysis model (MCDA)

Spatial prediction of Neural networks, urban Hybridized machine urban flood planning, GIS, artificial learning multi intelligence boosting (MultiB-MLPNN), MLPNN

Flood detection

Focus of the paper

Table 1 Analysis of the flood management techniques Research gaps/limitations

Model depends on local geo environmental dataset for implementations at different locations

Inclusion of hydraulic data such as inundation depth, velocity for hydraulics-based models

(continued)

Rainfall and water level Not determined the flood intensity

Curvature, distance from river, elevation, flow accumulation, river density, rainfall land cover and location of flood

Elevation, slope, precipitation, distance from river, Curve Number (CN) factor and distance from channel

Landmarks were roads, Inclusion of floodwater buildings, and bridges depth is underlined in the paper; can be used in determination of flood intensity in a region

Parameters analyzed

A Review on Urban Flood Management Techniques for the Smart City … 309

Author(s)

Wu et al. [7]

Naik et al. [30]

Yoon [10]

Rjeily [17]

S. No.

5

6

7

8

Table 1 (continued)

Urban flood forecasting

Urban flood forecasting

Flood forecasting

Urban flood depth prediction

Focus of the paper

Techniques used

Urban drainage systems, flooding forecast, NARX neural network, proactivity

Hydraulic modelling, short-term forecasting; blended QPF, floods

Flood prediction, machine learning, logistic regression

Nonlinear Auto Regressive with eXogenous inputs (NARX)

Adaptive blending method, radar-based extrapolation, real-time urban runoff forecasting using QPFs

Logistic regression model

Depth prediction, urban Regression model, flood, deep learning, deep learning, data warehouse Gradient Boosting Decision Tree (GBDT)

Key points in the coverage

Water depth variation, rainfall intensity

Radar-based image data, rainfall, 1059 pipes, 773 manholes, 772 sub basins

Rainfall

Water evaporation, rainfall duration, peak rainfall, catchment area, slope, roads, buildings, grassland

Parameters analyzed

(continued)

Can test for large dataset for flood prediction accuracy

Inclusion of diverse rainfall cases and study areas, utilizing a spatial mixing approach, weights are optimized using multiple storm events and criteria to increase forecast accuracy

Flood prediction accuracy is poor and need to be improve

Model relies on the conditional factors such as slope, land use, buildings not on the real time driving factors like heavy rainfall, high soil moisture, water velocity, volume etc

Research gaps/limitations

310 A. M. Hingmire and P. R. Bhaladhare

Author(s)

Lohani et al. [29]

Chen et al. [21]

Simoes et al. [22]

Mendoza-Cano et al. [32]

Chang and Chang [33]

S. No.

9

10

11

12

13

Table 1 (continued)

Urban flood warning system

Urban flood monitoring and warning

Urban drainage flood forecasting

Urban flood forecasting and warning

Real time flood forecasting system

Focus of the paper Fuzzy inference system, modified Takagi Sugeno (T–S) fuzzy inference system

Techniques used

Regional flood inundation forecast, Artificial Neural Networks (ANN), spatial–temporal distribution, Artificial Intelligence (AI)

Tropical storms, digital water network, early warning systems, IoT

Self-Organizing feature Mapping networks (SOM), Recurrent Nonlinear Auto Regressive with eXogenous inputs network (R-NARX)

Network design and sensor locations, smart water network

Pluvial flooding, flood Hydraulic 1D/1D, forecasting, hybrid 1D/2D model models, urban drainage models

Fuzzy evaluation, urban DPSIR model, flood forecasting DMFEW model model, fuzzy comprehensive evaluation, urban risk warning, radar based precipitation estimation

Self-Organizing Map (SOM), flood forecasting, fuzzy system

Key points in the coverage

It takes long simulation time

For flood forecasting and warning not focused on flood resilience method

Inclusion of flood warning system

Research gaps/limitations

Flood depth

(continued)

Inclusion of flood resilience method

Fluvial water level, soil Inclusion of flood moisture and weather resilience method parameters

Flow and water depth

Pressure index, impact index, force index, state index, and response index

Rainfall and water discharge data

Parameters analyzed

A Review on Urban Flood Management Techniques for the Smart City … 311

Author(s)

Vinothini and Jayanthy [34]

S. No.

14

Table 1 (continued)

Flood detection and notification system

Focus of the paper Sensor data, flood detection system, decision tree, IOT

Key points in the coverage Decision tree classifier

Techniques used Temperature, humidity and water levels

Parameters analyzed Can test for large dataset for flood detection accuracy

Research gaps/limitations

312 A. M. Hingmire and P. R. Bhaladhare

A Review on Urban Flood Management Techniques for the Smart City …

313

Many researchers have proposed various computational intelligence algorithms for flood management, such as fuzzy systems, neural networks, machine learning, CNN, SVM, hybrid algorithms are discussed in this paper. Different dataset used in flood forecasting and flood detection by the researcher is images, water level, Temperature, Humidity, soil moisture, rainfall, water depth, water flow, pressure, water discharge data etc. Ultrasonic, radar, and depth sensors are the three most often utilized sensors for IoT water depth measurements. The 2–400 cm non-contact measuring feature is offered by the ultrasonic ranging module HC-SR04, and the ranging precision is up to 3 mm [37]. The QHR104 is a radar water level sensor that measures levels continuously. The gadget measures the water level without coming into touch with the water by using the pulsed radar technique. It is examined that the most of research focuses on flood forecasting and monitoring from last two decades and proposes solutions based historical data and the real time data from flood forecasting to hydraulic models. The recent work mainly focuses on preparedness of the flood by prediction, monitoring and alerting models. The research gap analyzed of the discussed methodologies was the models tested by the researchers are on the smaller data size and accuracy of flood prediction variable. Table 2 shows the analysis of various flood forecasting techniques such as ANN, KNN, Logistic Regression, Deep Learning, Gradient Boosting Decision Tree, Fuzzy inference, SVM and hybrid models. Figure 2 shows that the accuracy of the most of models is between 80 and 90% but it is variable depending upon the cases and parameters. Though the prediction is above 76.48% in average, the rainfall and the flood is still uncertain. Hence the need arises to focus upon the development of flood resilience model which will automatically control or reduce the intensity of flood for the smart city.

4 Conclusion Building a real-time flood forecasting system requires knowledge from several academic disciplines. They include skills in wireless flood data transfer, sensor data fusion, data sensing at the appropriate time and place, model construction, remote weather station prediction, and flood resilience. New avenues have been opened up by IoT and computational models like the fuzzy model, artificial neural networks (ANNs), machine learning, and hybrid models, permitting the creation of new hardware and software for flood monitoring and forecasting. In order to reduce the impact of flood disasters through early warning, researchers and practitioners had conducted studies in the fields of collecting flood data, monitoring, forecasting, and detecting floods as well as early warning systems and data visualization. Many flood forecasting methods and hydrological models have been developed globally. The most of the research that have been studied at dealt with flood predicting, flood detection, and flood warning. In order to lessen flood-related fatalities, it is urgently necessary to create an efficient urban flood control system.

314

A. M. Hingmire and P. R. Bhaladhare

Table 2 Performance analysis of flood forecasting techniques S. No.

Author(s)

Techniques used

Accuracy

1

Hafiz et al. [1]

Haar cascade classifier, CNN classifier

91%

2

Darabi et al. [2]

Hybridized machine learning multi boosting (MultiB-MLPNN), MLPNN

26%

3

Samikwa et al. [6]

Artificial Neural Network (ANN), LSTM model

97.6%

4

Wu et al. [7]

Regression model, deep learning, Gradient Boosting Decision Tree (GBDT)

Relative error of prediction: 11.52%

5

Naik et al. [30]

Logistic regression model

75%

6

Rjeily [17]

Nonlinear Auto Regressive 99.8% with eXogenous inputs (NARX)

7

Lohani et al. [29]

Fuzzy inference system, modified Takagi–Sugeno (T–S) fuzzy inference system, SVM, ANN

TSC-T-S fuzzy model: 49% SC-T-S fuzzy model: 48.6% SOM: 48.5% ANN: 48.1%

8

Chen et al. [21]

DPSIR model, DMFEW model

80%

9

Fotovatikhah et al. [14]

ANN model

89%

10

Fotovatikhah et al. [14]

SVM-LN (linear) model, PL (polynomial), RBF (radial basis function), sigmoid (SIG)

89%

11

Sankaranarayanan et al. [36]

SVM model

85.57%

12

Fotovatikhah et al. [14]

k-Nearest Neighbour (kNN) model

87%

13

Tehrany et al. [35]

Decision Tree (DT)

87%

14

Fotovatikhah et al. [14]

Logical Regression (LR) model

73%

15

Fotovatikhah et al. [14]

Naïve Bayes model

72%

A smart water management solution and an autonomous urban flood control system may be constructed to lessen the effects of urban flooding in the smart city.

A Review on Urban Flood Management Techniques for the Smart City …

315

Fig. 2 Accuracy analysis of computational intelligence

References 1. Munawar HS, Ullah F, Qayyum S, Heravi A (2021) Application of deep learning on UAV-based aerial images for flood detection. Smart Cities 4:1220–1242. https://doi.org/10.3390/smartciti es4030065 2. Darabi H, Rahmati O, Naghibi SA, Mohammadi F, Ahmadisharaf E, Kalantari Z, Torabi Haghighi A, Soleimanpour SM, Tiefenbacher JP, Bui DT (2021) Development of a novel hybrid multi-boosting neural network model for spatial prediction of urban flood. Geocarto Int. http://doi.org/10.1080/10106049.2021.1920629 3. Feng B, Zhang Y, Bourke R (2021) Urbanization impacts on flood risks based on urban growth data and coupled flood models. Nat Hazards 106:613–627. http://doi.org/11069-020-04480-0 4. Indrasari W, Kadarwati LV. Prototype of water level monitoring system using magnetic sensor and ultrasonic based on Arduino Mega 2560. ICOSTA 2021. J Phys Conf Ser. http://doi.org/ 10.1088/1742-6596/2193/1/012052 5. Pham BT, Luu C, Van Dao D, Van Phong T, Nguyen HD, Van Le H, von Meding J, Prakash I (2021) Flood risk assessment using deep learning integrated with multi-criteria decision analysis. Knowl Based Syst 219:106899. 0950-7051/2021. http://doi.org/10.1016/j.knosys.2021. 106899 6. Samikwa E, Voigt T, Eriksson J (2020) Flood prediction using IoT and artificial neural networks with edge computing. In: 2020 international conferences on internet of things (iThings) and IEEE green computing and communications (GreenCom) and IEEE cyber, physical and social computing (CPSCom) and IEEE smart data (SmartData). http://doi.org/10.1109/iThings-Gre enCom-CPSCom-SmartData-Cybermatics50389.2020.00053 7. Wu Z, Zhou Y, Wang H, Jiang Z (2020) Depth prediction of urban flood under different rainfall return periods based on deep learning and data warehouse. Sci Total Environ. https://doi.org/ 10.1016/j.scitotenv.2020.137077 8. Kim HI, Han KY (2020) Urban flood prediction using deep neural network with data augmentation. Water 12:899. http://doi.org/10.3390/w12030899

316

A. M. Hingmire and P. R. Bhaladhare

9. Carlson K, Chowdhury A, Kepley A, Somerville E, Warshaw K, Goodall J (2019) Smart cities solutions for more flood resilient communities. In: IEEE Xplore 2019 systems and information engineering design symposium (SIEDS). http://doi.org/10.1109/SIEDS.2019.8735625 10. Yoon S-S (2019) Adaptive blending method of radar-based and numerical weather prediction QPFs for urban flood forecasting. Remote Sens 11:642. http://doi.org/10.3390/rs11060642 11. Qian K, Mohamed A, Claudel C (2019) Physics informed data driven model for flood prediction: application of deep learning in prediction of urban flood development. https://arxiv.org/abs/ 1908.10312 12. Kumar N, Agrawal A, Khan RA (2019) Cost estimation of cellularly deployed IoT-enabled network for flood detection. Iran J Comput Sci 2:53–64. http://doi.org/10.1007/s42044-01900031-4 13. Arshad B, Ogie R, Barthelemy J, Pradhan B, Verstaevel N, Perez P (2019) Computer vision and IoT-based sensors in flood monitoring and mapping: a systematic review. Sensors 19:5012. http://doi.org/10.3390/s19225012 14. Fotovatikhah F, Herrera M, Shamshirband S, Chau K, Ardabili SF, Jalil Piran Md (2018) Survey of computational intelligence as basis to big flood management: challenges, research directions and future work. Eng Appl Comput Fluid Mech 12(1):411–437. http://doi.org/10. 1080/19942060.2018.1448896 15. Keung KL, Lee CKM, Ng KKH, Yeung CK (2018) Smart city application and analysis: realtime urban drainage monitoring by IoT sensors: a case study of Hong Kong. In: Proceedings of the 2018 IEEE international conference on industrial engineering and engineering management (IEEM). http://doi.org/10.1109/IEEM.2018.8607303 16. Souza AS, de Lima Curvello AM, dos Santos de Souza FL, da Silva HJ (2017) A flood warning system to critical region. Procedia Comput Sci 109C:1104–1109 17. Rjeily YA, Abbas O, Sadek M, Shahrour I, Chehade FH (2017) Flood forecasting within urban drainage systems using NARX neural network. Water Sci Technol. http://doi.org/10.2166/wst. 2017.409 18. Zhang W, Wang X, Liu Y, Zhang T (2016) Simulation of rainstorm waterlogging based on SWMM and visualization module research. In: 2016 IEEE international conference on smart city and systems engineering. http://doi.org/10.1109/ICSCSE.2016.28 19. Rothkrantz LJM (2016) Flood control of the smart city Prague. In: IEEE smart cities symposium Prague 2016. http://doi.org/10.1109/SCSP.2016.7501043 20. Lo S-W, Wu J-H, Lin F-P, Hsu C-H (2015) Visual sensing for urban flood monitoring. Sensors 15:20006–20029. http://doi.org/10.3390/s150820006. ISSN 1424-8220 21. Chen Y, Zhou H, Zhang H, Du G, Zhou J (2015) Urban flood risk warning under rapid urbanization. Environ Res 139:3–10. www.elsevier.com/locate/envres. http://doi.org/10.1016/j.env res.2015.02.028 ˇ (2011) Urban drainage 22. Simões N, Ochoa S, Leitão JP, Pina R, Sá Marques A, Maksimovi´c C models for flood forecasting: 1D/1D, 1D/2D and hybrid models. In: 12th international conference on urban drainage, Porto Alegre/Brazil, 11–16 Sept 2011 23. Yusoff A, Mustafa IS, Yussof S, Din NM (2015) Green cloud platform for flood early detection warning system in smart city. In: 2015 5th national symposium on information technology: towards new smart world (NSITNSW). IEEE. http://doi.org/10.1109/NSITNSW.2015.7176406 24. Xu G, Huang GQ, Fang J, Qiu X (2014) An integrated cloud platform for cooperative smart asset management in urban flood control. In: Proceedings of the 2014 IEEE 18th international conference on computer supported cooperative work in design. http://doi.org/10.1109/ CSCWD.2014.6846820 25. Narayanan RK, Lekshmy VM, Rao S, Sasidhar K. A novel approach to urban flood monitoring using computer vision. In: Fifth international conference on computing, communications and networking technologies (ICCCNT). IEEE. http://doi.org/10.1109/ICCCNT.2014.6962989 26. Sayers W, Savi´ca D, Kapelan Z, Kellagher R (2014) Artificial intelligence techniques for flood risk management in urban environments. In: 12th international conference on computing and control for the water industry, CCWI 2013. Procedia Eng 70:1505–1512. http://doi.org/10. 1016/j.proeng.2014.02.165

A Review on Urban Flood Management Techniques for the Smart City …

317

27. Sunkpho J, Ootamakorn C (2011) Real-time flood monitoring and warning system. Songklanakarin J Sci Technol 33(2):227–235 28. Bruen M, Yang J (2006) Combined hydraulic and black-box models for flood forecasting in urban drainage systems. J Hydrol Eng. ISSN 1084-0699/2006/6-589 29. Lohani AK, Goel NK, Bhatia KKS (2014) Improving real time flood forecasting using fuzzy inference system. J Hydrol 509:25–41. https://doi.org/10.1016/j.jhydrol.2013.11.021 30. Naik S, Patil SA, Verma A, Hingmire A (2020) Flood prediction using logistic regression for Kerala state. Int J Eng Res Technol (IJERT) 09(03) 31. https://www.smartcitygovt.com/blog/2018/8/21/flooding-and-the-smart-city 32. Mendoza-Cano O, Aquino-Santos R, López-de la Cruz J, Edwards RM, Khouakhi A, Pattison I, Rangel-Licea V, Castellanos-Berjan E, Martinez-Preciado MA, Rincón-Avalos P, Lepper P, Gutiérrez-Gómez A, Uribe-Ramos JM, Ibarreche J, Perez I. Experiments of an IoT-based wireless sensor network for flood monitoring in Colima, Mexico. J Hydroinformatics 23(3):385. http://doi.org/10.2166/hydro.2021.126 33. Chang L-C, Chang F-J (2020) IoT-based flood depth sensors in artificial intelligent urban flood warning systems. In: EGU general assembly 2020. https://doi.org/10.5194/egusphereegu2020-12523 34. Vinothini K, Jayanthy S (2019) IoT based flood detection and notification system using decision tree algorithm. In: Proceedings of the international conference on intelligent computing and control systems (ICICCS 2019). IEEE Xplore Part Number: CFP19K34-ART; ISBN: 978-15386-8113-8 35. Tehrany MS, Pradhan B, Jebur MN (2013) Spatial prediction of flood susceptible areas using rule based decision tree (DT) and a novel ensemble bivariate and multivariate statistical models in GIS. J Hydrol 69–79. http://doi.org/10.1016/j.jhydrol.2013.09.034 36. Sankaranarayanan S, Prabhakar M, Satish S, Jain P, Ramprasad A, Krishnan A (2020) Flood prediction based on weather parameters using deep learning. J Water Clim Change. http://doi. org/10.2166/wcc.2019.321 37. https://cdn.sparkfun.com/datasheets/Sensors/Proximity/HCSR04.pdf

Application of Distributed Constraint Optimization Technique for Privacy Preservation in Cyber-Physical Systems Manas Kumar Yogi and A. S. N. Chakravarthy

Abstract As the modern world is applying smart technologies in every day and in every sphere of life, the advent of Industry 4.0 standards increase the risk of privacy loss while using the cyber physical systems. There are many popular methods of privacy preservation in CPS but they are too complex to implement and not have high degree of data utility. In keeping all these reasearch challenges, we have applied a novel approach of distributed constraint optimization for minimizing the privacy loss in a cyber physical system. The main advantage of our proposed mechanism is that the balance between local privacy level and global privacy level is maintained so that the data utility is not degraded. Keywords Cyber physical systems · Privacy · Distributed · Utility · Constrainst

1 Introduction Network and interoperability have become critical with regards to the digitized business, being liable for associating items, machines, individuals and the climate in a solitary shrewd assembling system. The reconciliation of these with data advancements has driven organizations to accomplish levels of proficiency previously unheard of. Data innovations support the capacity, assortment and investigation of information while working innovations support the production of actual value. It is the consolidation between the two that leads to the much sought-after ‘cyberphysical systems’, made conceivable by a focal framework that controls and screens tasks at all levels [1]. Cyberphysical systems are a bunch of intuitive systems upheld by clever machines which, composed and constrained by a focal substance, send functional data to qualified laborers. Along these lines, creation tasks run under M. K. Yogi (B) CSE Department, Pragati Engineering College (Autonomous), Surampalem, Andhra Pradesh, India e-mail: [email protected] A. S. N. Chakravarthy CSE Department, JNTUK, Kakinada, Andhra Pradesh, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_24

319

320

M. K. Yogi and A. S. N. Chakravarthy

restraint while engineers deal with on expected issues and potential arrangements straightforwardly in this robotic copy. In numerous areas of human exercises, CPS has acquired and more consideration, particularly in the limits where physical cycles and physical hardware should have been controlled, arranged and facilitated smartly. Taking into account the intricacy of present day CPS, issues of guaranteeing the security and wellbeing of those systems are of high pertinence [2, 3]. The potential dangers can be connected with a cyber, physical or the two components of CPS and in this manner require complex methodology for recognizable proof and relief of safety and security weaknesses. Distributed constraint optimization (DCOP or DisCOP) is the distributed simple to constraint optimization [4, 5]. A DCOP is an issue where a gathering of agents must distributedly pick values for a bunch of variables to such an extent that the expense of a bunch of constraints over the variables is limited. Distributed Constraint Satisfaction is a framework for depicting an issue as far as constraints that are known and enforced by particular members (agents). The constraints are portrayed on certain variables with predefined spaces, and must be appointed to similar qualities by the various agents. Issues characterized with this framework can be tackled by any of the algorithms that are designed for it. DCOPs are a well known approach to formulating and taking care of multi-specialist coordination issues, for example, the distributed scheduling of gatherings, distributed coordination of automated air vehicles and the distributed designation of focuses in sensor networks [6, 7]. Privacy worries in the scheduling of gatherings and the limit of communication and calculation assets of every sensor in a sensor network makes brought together constraint optimization troublesome. Therefore, the idea of these applications require a distributed methodology.

2 Related Work With the inescapability of cell phones and situating advances, area based administrations (LBSs) turned out to be progressively normal in practically all friendly and business spaces [8, 9]. Various portable LBS applications, for example, route, long range interpersonal communication and administration suggestion have been formed and incorporated into individuals’ day to day exercises, giving accommodating information about their environmental factors. Notwithstanding, when trajectory information are distributed for examination, a pernicious client might have the option to find individual and delicate information of people, regardless of whether any expressly recognizing information is eliminated before distributing [10]. With the information on semi identifiers, i.e., focuses that can be connected to outer information and used to reidentify people, an assailant might have the option to anticipate back the mysterious individual development information [11]. Hence, extreme worries about privacy are raised, since this measure of information gives area information that recognizes people and, possibly, their touchy information like social traditions, strict inclinations and sexual inclinations. In the recent times, the course

Application of Distributed Constraint Optimization Technique …

321

of information distribution has expanding and more perplexing. The extents of informational indexes have developed exorbitantly to where the adequacy of customary algorithms, which run in concentrated registering conditions, is turning out to be progressively troublesome. It is normal to handle enormous informational indexes with the guide of distributed platforms, for example, the MapReduce framework, to work with distributed information handling for huge scope bunches and accomplish significant performance improvement [12]. Information anonymization approaches proposed for incorporated figuring conditions normally present poor performance to process identifiers and to offer privacy from this information. Furthermore, they have zeroed in on the nature of privacy protection and the utility of the distributed information. Notwithstanding, adaptability is additionally an important issue when the quantity of semi identifiers are excessively high. Therefore, the fundamental inquiry raised while thinking about this multitude of issues is: “Is it conceivable to anonymize huge scope informational collections without losing adaptability?” The research lies mainly in the aspects of dynamical systems like CPS where privacy cannot be treated as a centralised feature [11]. Due tot he privacy requirements of users changing from time to time and the uncertainty in the way they interact with an CPS entity, forced us to think in the aspect of distributed privacy [13]. It is easier to postualte company privacy policies for an organisation but while enforcing the rules these consume more effort and time. Hence the main challenge lies in optimizing the privacy requirements by formulating it as a distributed constraint optimization issue. Many advances in privacy models have been made to tackle the issue of information anonymization. k-secrecy is a notable model zeroed in on social data set conspires, that forestalls character revelation by guaranteeing that for each mix of upsides of semi identifiers, there are basically k records in the distributed informational index [14]. All in all, each record can not be linked to a person by an attacker with likelihood lower than 1/k. K-secrecy has likewise been as of late considered by trajectory anonymization strategies to offer privacy mindful information publishing, although it has been shown that it presents a few impediments and finding an ideal k-anonymization is NP-hard. Speculation and concealment are the most utilized strategies to ensure the k-obscurity property. In this situation, a few anonymization strategies have been proposed to safeguard trajectory against personality revelation. Never Walk Alone (NWA) utilizes (k, δ)-obscurity privacy model to anonymize trajectories by producing round and hollow volumes of sweep δ that contain basically k trajectories [15]. For this reason, trajectories are first isolated into disjoint gatherings grouped utilizing the Euclidean distance. Each group will contain basically k trajectories with Euclidean distance δ. Be that as it may, the algorithm might bring about little gatherings with less than k trajectories, not giving significant privacy ensures, practically speaking. scientists have additionally proposed a methodology in light of information speculation to make anonymization gatherings. They considered timestamps as semi identifiers and expected that every trajectory has its own arrangement of times as its semi identifier. Subsequently, they proposed an idea of k-namelessness for trajectory information by characterizing an attack graph for foes. In the event that each hub in the graph has degree k and the attack graph is symmetric, it implies a moving item fulfills k-namelessness. To accomplish the

322

M. K. Yogi and A. S. N. Chakravarthy

k-secrecy property for every trajectory, anonymization bunches are not be guaranteed to disjoint. In that work, the authors accept that the information distributer knows the semi identifiers for every trajectory, except they don’t specify how this knowledge can be acquired. As of late, one more category of methods, specifically differential privacy, has gotten a developing interest in information anonymization. Most works on differential privacy center around unambiguous information insightful tasks, for example, count inquiry addressing and continuous example mining. The primary thought of these strategies is to deliver loud information and to guarantee that the expulsion or expansion of a solitary record doesn’t fundamentally affect the result of any investigation. In this manner, unlike our work, they can’t protect information honesty, which is important to safeguard in numerous applications. Databases assume a critical part in logical movement. However, the actual fuel of databases, open data, behaves like a soil when matched with the meshwork of our security regulations. Since the beginning of reidentification a precarious circumstance has arisen: however we can assemble and handle extraordinarily tremendous measures of data, the legitimate sloughs encompassing sharing this data acts restrictively [16]. Synthetic data offers progress. However not a silver shot, the technique permits us to stop the deidentification-reidentification weapons contest and spotlight on what makes a difference, helpful data. To this degree, we suggest that the security local area acknowledge synthetic data as a legitimate, following stage to the database-protection issue. Our mechanism should be used in conjnction with the features of synthetic data to overcome the privacy leak pertinent in the most popular privacy preservation systems employed in CPS.

3 Proposed Mechanism We intend to apply the privacy preservation problem as a distributed constraint optimization problem. As the various entities in a CPS ecosystem interact with each other to complete their full functionality, the motivation is suitable. Also the privacy aspect is intended not to be a global issue but distributed local issues. In the sense, we apply the distributed constraint to the privacy of each CPS agent calling it as local privacy. When the agents send messages to other agents in the CPS ecosystem, they are aware of only thier local privacy level which must be optimized. Like this, we formulate the global privacy limit as an optimization problem as shown below. Consider the following optimization problem min

N 

P(x)

(1)

i=1

where P(x) represents the minimum privacy leak for the whole CPS ecosystem where the users are from 1, 2, 3… to N. These are legititate users. They want their sensitive information to be hidden without effecting the data utility.

Application of Distributed Constraint Optimization Technique …

323

We consider a network of N processors communicating according to a connected and undirected graph G = ((1, 2, 3, …N); E), where E denotes the set of edges. Edge (i, j) depicts the notion that node i sends data to j. Note that, being the graph undirected, for each (i, j) belongs to E, then (j, i) also belongs to E. We denote by [E] as the cardinality of E. Each CPS node i knows only fi , gi and Xi , and tries to contributes its privacy leak value at minimum level by acting locally. Distributed Constraint Algorithm for Privacy Preservation CPS node states: xi , αi , ϕi , μij for i, j belongs to 1, 2, …, N. Where xi is the privacy vector values of a CPS node i. αi represents the privacy barrier, i.e. maximum privacy loss that is permissible when two CPS nodes exchange information among themselves ϕi , μij are the auxilliary variables which enter into the privacy preservation scenario when scaled globally in the CPS ecosystem. Evolution: Collect μij from i, j Compute((xi, , αi ), ϕi ) as a primal dual solution pair of min

N 

P(x)

i=1

subject to αi > 0, xi belongs to E and

N  

 μij − μji ≤ 1

(2)

j=1

Collect ϕi from j belongs to E. Update for all values of j as shown below   t+1 t − ϕjt+1 μt+1 ij = μij − η ϕi

(3)

We can observe that, each CPS entity i tries to determine the optimal mechanism by interacting locally with neighbouring CPS entity. In the above proposed algorithm, the CPS nodes interact with each other by exchanging their privacy requirements but they donot reveal any information about their exact degree of privacy release. So, the feature is so robust that it leads to minimum privacy loss in the whole CPS. We have kept the initialisation of the privacy needs of the nodes at random values, so that the algorithm can handle the dynamic nature of privacy requirements of the inherent CPS nodes. If any CPS node join or leave the CPS ecosystem, the privacy preservation problem will be converted into a new optimization problem. Our novel approach will help in giving a solution which will eventually converse to a solution of the latest optimization problem.

324

M. K. Yogi and A. S. N. Chakravarthy

4 Experimental Results For deploying our proposed approach we have considered a lung cancer dataset downloaded from Kaggle which has 15 features and 284 instances. We have used python language to develop the code applying the distributed constraint optimization algorithm. In Fig. 1 it has been shown that our proposed novel mechanism performs better when compared to other popular methods of privacy preservation in a CPS ecosystem like location privacy, personal differential privacy, differential privacy. From the graph, we can observe that our proposed appraoch maintains a balanced level between data privacy and data utility. The main challenge is the tradeoff between data privacy and data utility and the distributed constraint optimization method tries to balance this tradeoff evenly. As evident from the graph below, the privacy loss reduces below 4% and data utility is nearly 10% with application of our proposed approach. From Fig. 2 we can observe how the increase in number of iterations helps in reduction of privacy loss. In this sceanario also, our proposed technqiue outperforms the other three popular methods of privacy preservation in a cyber physical system. As observed fromt the above graph, with nearly 10 iterations the privacy loss can be brought down below 4% but other methods have privacy loss of nearly 5% on avaerage even with 10 iterations of the algorithmic run. From Fig. 3,we can infer that as the number of interactions between the various CPS nodes increases, the local as well as global privacy level of the entire CPS also increases. By using our novel approach the privacy level increases by nearly 7% which is better than other methods of privacy preservation in comparision. The global level of privacy may remain below the accepted level of privacy which is

Fig. 1 Plot of data utility versus privacy loss among popular methods for privacy preservation in CPS

Application of Distributed Constraint Optimization Technique …

325

Fig. 2 Impact of number of iterations of the methods against privacy loss (%)

less than the worst case privacy level of sum of all local privacy levels of individual CPS nodes. Whether it is a cyber node or physical node, the privacy leak should be kept at a minimum level else the auxilliary variables values in the algorithm cannot decrease the privacy levels. The auxilliary variables help in keeping the local privacy levels bounded. This helps in scaling the overall global privacy level with a minimum cost function. The local agents donot know what is the global privacy level and this principle of abstraction supports in enhancing the total data utility value at optimum level. This is the most appreciable fact about the proposed technique.

5 Future Work In problem formulation we have assumed few factors. The first assumption is that each inetracting agent in the CPS ecosystem is aware of a privacy constraint value and factors effecting the privacy constraints. But in reality, it may not be the case. The second assumption is that, each agent interacts with its neighboring agents only by passing a direct signal. We are ruling out indirect communication between the CPS nodes and various agents but again the doesnot hold true in all the cases in real time. In a dynamic system like CPS, there are multiple instances of indirect communication between the cyber components and physical components. In future work, we will try to reduce the scope of this assumptions so that the robustness of our proposed approach will increase and the dependency of the agents on the indirect interactions between the various CPS nodes will not effect much the inherent privacy loss. In future we also want to develop privacy preservation models. In future we also plan to introduce synthetic data as input to the proposed technqiue so that there

326

M. K. Yogi and A. S. N. Chakravarthy

Fig. 3 Plot of number of inetractions between various CPS nodes versus local and global privacy levels (in %)

is a two-fold benefit in privacy level. We want to advocate a factor of privacy barrier which will act as an indicator of how the balance between local privacy level and global privacy levels can be optimized so that data utlity is maximized. Also, there lies oppurtunity to include the adversary knowldege to develop threat models on which the proposed techniques should be deployed so as to know the weaknesses, if any on our novel approach. Overall, we can say that in future lot of potential design issues in privacy can be solved by working in this future direction.

6 Conclusion Our paper will serve as a guide for researchers who want to explore the constraint satisafction strategy for optimal allocation of privacy budget in each participating node of a cyber-physical system. The distributed constraint optimization helps in maintaining the overall privacy loss. With these approach the balance between data privacy and data utility is also maintained because if it is not maintained then the user will not get satisafction with the functionality of the system. This paper provides a robust framework on which future researchers can build efficient privacy preservation models by extending with other types of distributed constraint satisafaction models.

Application of Distributed Constraint Optimization Technique …

327

References 1. Ali W et al (2020) ALPHA: an anonymous orthogonal code-based privacy preserving scheme for ındustrial cyber–physical systems. IEEE Trans Ind Inform 17(11):7716–7724 2. Keshk M et al (2021) Privacy-preserving schemes for safeguarding heterogeneous data sources in cyber-physical systems. IEEE Access 9:55077–55097 3. Kanchan S, Singh G, Chaudhari NS (2022) SPSR-VCP: secure and privacy preserving SignRecryption in vehicular cyber physical systems. J Ambient Intell Humaniz Comput 13(1):1–20 4. Cai Z, Zheng X (2018) A private and efficient mechanism for data uploading in smart cyberphysical systems. IEEE Trans Netw Sci Eng 7(2):766–775 5. Yu J et al (2018) Privacy-preserving data aggregation computing in cyber-physical social systems. ACM Trans Cyber Phys Syst 3(1):1–23 6. Feng J et al (2020) Privacy-preserving computation in cyber-physical-social systems: a survey of the state-of-the-art and perspectives. Inf Sci 527:341–355 7. Bhattacharjee A et al (2021) Vulnerability characterization and privacy quantification for cyberphysical systems. In: 2021 IEEE international conferences on internet of things (iThings) and IEEE green computing and communications (GreenCom) and IEEE cyber, physical and social computing (CPSCom) and IEEE smart data (SmartData) and IEEE congress on cybermatics (Cybermatics). IEEE 8. Wang H, Fan K, Zhang K, Wang Z, Li H, Yang Y (2021) Secure and efficient data privacypreserving scheme for mobile cyber physical systems. IEEE Internet Things J 9. Kanagaraj A et al (2022) Differential privacy techniques-based ınformation security for cyber physical system applications: an overview. In: Real-time applications of machine learning in cyber-physical systems, pp 52–64 10. Lu W et al (2021) Edge blockchain assisted lightweight privacy-preserving data aggregation for smart grid. IEEE Trans Netw Serv Manag 18(2):1246–1259 11. Gao C et al (2021) Sampled-data-based fault-tolerant consensus control for multi-agent systems: a data privacy preserving scheme. Automatica 133:109847 12. Li Q et al (2021) Privacy-preserving distributed processing: metrics, bounds and algorithms. IEEE Trans Inf Forensics Secur 16:2090–2103 13. Hefner T, Shani G, Stern R (2022) Privacy preserving planning in multi-agent stochastic environments. Auton Agent Multi-Agent Syst 36(1):1–27 14. Wang Y et al (2021) Privacy-preserving consensus for multi-agent systems via node decomposition strategy. IEEE Trans Circuits Syst I Regul Pap 68(8):3474–3484 15. Mugunthan SR (2019) Security and privacy preserving of sensor data localization based on internet of things. J ISMAC 1(02):81–92 16. Shakya S (2019) Efficient security and privacy mechanism for block chain application. J Inf Technol 1(02):58–67

Grip Assisting Glove for Charcot-Marie-Tooth Patients Varun Sarathchandran, Jason Vincent, Juel Mathais George, Polu Sathwik Reddy, and R. Ambika

Abstract Charcot-Marie-Tooth is a neurodegenerative disease which causes muscle degeneration and loss of sense in the appendages. Patients affected with the disease find holding an object in their hand difficult, since they have no sensation in their skin and have no idea how well they are grasping it. Over time, it is possible that their grip falls, and the object slips from their hand. To counter this, a device which improves the quality of life of people affected with Charcot-Marie-Tooth has been presented in this paper. The proposed device is in the form of a glove, which uses force sensors to continuously track the force applied by the user on the object they are holding. If it falls below a certain threshold, it warns them through haptic feedback-vibration motors placed in the forearm where the users’ sensation of touch is maintained. The sensors are paired with analog signal conditioning circuitry to obtain as good a linearity as possible without sacrificing resolution. Keywords Assistive technology · Analog signal processing · Force measurement · Haptic feedback

1 Introduction Charcot-Marie-Tooth (CMT) is a neurodegenerative disease that leads to the degeneration of a neuron’s myelin sheath which causes poor sensation in the patient’s fingers and toes. Methods to rehabilitate the hand function of patients are scarce since CMT is a rare disease, and studies of how hand function decreases over disease progression is sparse [1, 2]. Charcot-Marie-Tooth affects approximately 1 in 3000 people [3]. It is the most common inherited neuro-muscular disorder [4]. The severity of symptoms can vary greatly from person to person. It is passed on genetically and is caused by mutations in the genes that affect the nerves in your extremities [5, 6]. Daily activities become difficult tasks for those affected-owing to weakness in the V. Sarathchandran (B) · J. Vincent · J. Mathais George · P. Sathwik Reddy · R. Ambika BMS Institute of Technology and Management, Yelahanka, Bangalore 560064, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_25

329

330

V. Sarathchandran et al.

hands along with decreased sensitivity to touch, heat, and cold in the feet and lower legs. This project aims to design a system that detects and informs the user affected by CMT when their grip on an object recedes. A device is offered in the form of a glove which continuously measures the amount of force the user is applying on something they are holding. When the device senses that the user is gradually losing their grip, a vibration motor triggers in the forearm- indicating to the user that they must reinforce their grip. The system makes use of a set of force sensors that sends signals through an analog signal conditioning circuit which is then fed to the microcontroller. The microcontroller uses the conditioned signal and determines if the received signal crosses the defined threshold required to trigger the vibration motor. The threshold defined is a marker that indicates when the force detected by the force sensor is dropping, that is, it indicates if the person holding an object is losing their grip. If the force signal drops below the threshold, the microcontroller sends the signal that initiates the vibration motor. The vibration motor located at the forearm, indicates the user that their grip is decreasing beyond the set threshold.

2 Literature Survey An extensive literature survey has been carried out and this section presents some of the existing methods.

2.1 Smart Muscle Strength Assessment Glove for Rehabilitation Purposes [7] A low cost, muscle strength assessment device was designed. It used five force sensors and a load cell to measure the force application capability of the muscles in the finger. The device then classified the users muscle function using Neural Networks. Real data was used to train the algorithm. Based on the result, the physiotherapist was advised a therapeutic plan.

2.2 Calibration and Evaluation of a Force Measurement Glove for Field-Based Monitoring of Manual Wheelchair Users [8] A glove was designed to detect loading events and relative changes in the application of force in wheelchair users. The glove comprises of four force sensors in the palm. Each FSR was fitted into the glove using sewed compartments. FSRs were connected

Grip Assisting Glove for Charcot-Marie-Tooth Patients

331

to a PCB through a cloth sleeve. An elastic band fastened the PCB housing to the user’s forearm. The PCB sampled the sensor, and the data was stored in a 32 GB microSD card for further processing. The glove managed to detect 72.7% of the peaks. A certain calibration condition, which was the average of the 5 conditions was tested, and gave the least error for all metrics and a linear correlation r equal to 0.80. Thus, it can be concluded that the average loading condition located at the palm is significantly aligned with the average of all the testing conditions than any individual condition.

2.3 A Fabricated Force Glove that Measures Hand Forces During Activities of Daily Living [9] 14 flexiforce sensors were used to measure forces applied by different parts of the fingers. Results from the study were used to determine minimal forces of the hand during activities of daily living to appropriately design exoskeletons and prosthetics. The flexiforce sensors—A201, which were attached to the palmar side of the hand, were directly connected to an arduino microcontroller via a resistive voltage divider circuit. The difference in voltage across one of the resistors was used to map the voltage-force ranges. A low pass filter of 10 Hz was then used to remove noise. The study concluded that forces applied during daily activities lie between 9 and 24 N.

2.4 Wearable System with Embedded Force Sensors for Neurological Rehabilitation Trainings [10] A glove which is compact and energy efficient, thus suitable for daily wearing by patients was designed. It integrated four A301 sensors on the four fingers. The device used a ATmega 328 on Bluno Beetle board capable of wireless transmission of signals. Since the sensor in itself is flexible, implementing it on a soft surface leads to inaccurate results. Thus, two small discs were 3D printed and adhered to both sides of the sensor head. A circuit for analog signal conditioning was designed to compromise in terms of force sensing resolution and linearity of transduction. The glove was designed to be used for neurological rehabilitation training.

2.5 Tactile Sensorized Glove for Force and Motion Sensing [11] The paper aimed to assess hand related injuries such as thumb tendonitis and carpal tunnel syndrome. It focused on the constant use of smartphones causing thumb and

332

V. Sarathchandran et al.

wrist injuries, which left untreated might cause permanent injury. Finger function testing methods are often bulky and not dynamic which restricts hand movements. To improve present devices, a glove system was developed which is tactile and sensorized. The system has several thin-film flexible and stretchable strain gauges and force sensors to independently measure finger movement and pressure. It provides real time accurate results within minutes and causes minimal discomfort to the wearer.

2.6 Smart Glove to Measure a Grip Force of the Workers [12] The aim of the project is to analyse the hand for musculoskeletal injuries that are a result of everyday activities or from repetitive loading in poor conditions. The prototype of the device contains 14 connected sensors. An electronics control unit on the wrist performed analog signal processing. The sensor current was amplified. Sensitivity of the sensor can be fine-tuned by manipulating the value of the feedback resistor. 12-bit successive approximation ADCs were used for digitization. Communication to the PC was done through a Bluetooth module.

2.7 Force-Sensing Glove System for Measurement of Hand Forces During Motorbike Riding [13] The design includes flexi-sensors on a glove, a chipboard, a Bluetooth trans-receiver and a smartphone. Two Flexi-Force A401 sensors were placed on the index and middle fingers to monitor the two-finger braking. Another two sensors were positioned on the thumb and palm to observe the rotation of the throttle. The microcontroller is the MCF51JM128VLH which is a 32-bit RISC device, and an ultra-low power MC. Bluetooth technology was used for real-time data communication. An algorithm was designed to identify the specific hand movements observed during riding a bike. Taking into account the force sensing elements, using the algorithm it was capable of differentiating between the actions of clutch, wheel, brake and throttle.

2.8 Smart Tactile Gloves for Haptic Interaction, Communication and Rehabilitation [14] These gloves utilise different types of sensors to measure the bending angle, pressure, and/or different orientations of the hand and/or fingers during interaction. The three main categories of smart gloves:

Grip Assisting Glove for Charcot-Marie-Tooth Patients

333

Gesture-based: These use cameras to capture hand movement. The videos are then processed using computer vision to determine function. Heavy cameras cause poor mobility and distress the wearer. Touch-based: Information is either sent or received by the user through touch using touch sensors. Gesture and Touch based: Uses a combination of both aforementioned methods.

3 Working/Methodology The basic block diagram of the proposed method is shown in Fig. 1. The device is considered to be made of three separate blocks. 1. Sensing and analog signal condition block 2. Continuous force monitoring block 3. Haptic feedback block.

3.1 Sensing and Analog Signal Condition Block Two A101 [15] sensors (sensor specifications in Fig. 2) are placed on the thumb, and the ring finger to continuously monitor the force applied by the hand onto an object held in it. The force sensors are the smallest offered by the firm FlexiForce and are of dimensions 7.6 mm × 15.6 mm. The sensing area is a circular disk of diameter 3.8 mm. The sensor can measure a maximum force of up to 10 pounds. It is a standard piezoresistive force sensor with a typical performance curve. The challenge is to design an analog signal conditioning system to accurately measure the force applied, with reasonable linearity in the requisite range. A standard potentiometer arrangement is found to be unusable because of extreme nonlinearity

Fig. 1 A representation of the basic block diagram of the proposed model

334

V. Sarathchandran et al.

Fig. 2 Signal and noise samples from the ring finger, and power spectrum

in the range. Thus, the given circuit is used. The circuit is a combination of the op amp MCP6004. The op amp is used in the inverting amplifier mode with a capacitor as feedback to perform noise cancellation. The feedback resistance, capacitor value and V reference values are calculated to specific values for the need. Sensor specifications are mentioned in Table 1. Circuit Design The value of Vref is maintained as −3.7 V, the output of a 3. 7 V LiPo battery which is used to provide for an isolated supply. The value of the feedback resistor is chosen to be the same value of that of the sensor resistance at maximum force applied when a user grabs onto an object. This ensures that the input to the microcontroller will be about 3.7 V when an object is normally held, ensuring both a leeway of 1.3 V for unpredictable signal swing, as well as sufficient resolution when the force on the object begins to fall. The capacitor in the feedback acts as a single pole low pass filter of cut off frequency 1/2π RC

(1)

The spectrum of noise is studied on an oscilloscope and MATLAB to determine the requisite cut off frequency. The capacitor value is then calculated. An additional single pole LPF is considered at the output of the op amp, but is not implemented

Grip Assisting Glove for Charcot-Marie-Tooth Patients

335

Table 1 Sensor specifications from manufacturer Typical performance

Evaluation conditions

Linearity (Error)

< ±3% of full scale

Line drawn from 0 to 50% load

Repeatability

< ±2.5%

Conditioned sensor, 80% of full force applied

Hysteresis

Ftarget , Compare the findings of the cascaded detector to the collection of non-face photos, and add any false positives to the set Q. Hardware Components Used ESP32, which is integrated with 802.11b/g/n Wi-Fi and Bluetooth in dual mode (BT). Jumper wires is used for connecting purposes. The battery with cap is used to power BO motors so they may continue to move. Additionally, it serves as a power

362

U. Sahana and N. Rajesh

supply for the submersible motor that activates the sanitizer dispenser. It can hold 9 V. BO Motors are 100 RPM DC motors that aid in moving the ROBOT model in the desired directions. Wireless temperature sensor MLX90614, it is used to measure body temperature of the patients, it is interfaced to ESP32 board, soon after face of patient is scanned using pi cam, temperature is displayed on the monitor. The Raspberry Pi 3 Model B+ has been used with a camera module attached to it (Figs. 4, 5 and 6). Fig. 4 Robot

Fig. 5 Webcam and temperature sensor

Wheeled Robots for Isolation Wards

363

Fig. 6 Dashboard showing body temperature of patients

5 Results and Discussion 6 Conclusion The proposed model is an attempt to leverage the robot and internet of things concepts to create an innovative product that is useful and useful to society to tackle situations like Covid 19, and it is aimed at frontline health warriors. The prototype has been evaluated for a variety of test scenarios and has functioned well. However, it does have some limitations, which are stated in the document’s drawbacks section. The prototype could be further enhanced and made usable in real-time applications by considering the restrictions indicated in the preceding sections. Applications • It can be used in isolation wards, intensive care units, and quarantine centers to keep track of patients or infected people from afar. • Helpful for supplying food and medicine to pandemic victims or suspects with little or no human intervention. • It allows for the retrieval of a patient’s temperature from a remote location without the need for human intervention. • A machine learning model obtains the patient’s identity and enters the patient’s name and body temperature into a database after scanning an image through a camera.

364

U. Sahana and N. Rajesh

References 1. Akhund TMNU, Jyoti WB, Siddik MAB (2020) IoT based low-cost robotic agent design for disabled and covid-19 virus affected people. In: 2020 Fourth world conference on smart trends in systems, security and sustainability (WorldS4) 2. Antony M, Parameswaram M, Mathew N, Sajithkumar VS, Joseph J, Jacob CM (2020) Design and implementation of automatic guided vehicle for hospital application. In: Proceeding of the fifth international conference on communication and electronics systems (ICCES 2020) 3. Kalaiarasi D, Pavithra S, Pratheeba S, Priyaadharshini RL (2018) IoT based motion control system of a robotic car. Int Res J Eng Technol (IRJET) 05(03). e-ISSN: 2395-0056 4. Durga Devi S, Mounasri P, Mounika S (2017) IoT based remote controlled robot using android. IJESC 7(3) 5. Shyla R, Megha Shree AC (2017) Intelligent wheel chair based on internet of things. Int Res J Eng Technol (IRJET) 04(08). e-ISSN: 2395-0056 6. Duseja L, Deshmukh Y, Karmuse S, Ohol SS (2021) Autonomous RFID controlled assisting robot for isolation wards. In: Proceedings of 6th international conference on intelligent technologies (ICIT-2021), vol 17, p 19 7. Prabhakar M, Paulraj V, Dhanraj JA, Nagarajan S, Kannappan DAK, Hariharan A (2020) Design and simulation of an automated guided vehicle through webots for isolated COVID-19 patients in hospitals. In: 2020 IEEE 4th conference on information and communication technology (CICT). IEEE, pp 1–5 8. Choi HK, Cui C, Seok H, Bae J-Y, Jeon JH, Lee GE, Choi WS, Park M-S, Park DW (2022) Feasibility of ultraviolet light-emitting diode irradiation robot for terminal decontamination of coronavirus disease 2019 (COVID-19) patient rooms. Infect Control Hosp Epidemiol 43(2):232– 237 9. Yogesh S, Prasanna B, Parthasarathi S, Ganesh MA (2020) Open MV-micro python based DIY low cost service robot in quarantine facility of COVID-19 Patients. In: Congress on intelligent systems. Springer, Singapore, pp 519–531

A Survey on Various Crypto-steganography Techniques for Real-Time Images R. Tanya Bindu and T. Kavitha

Abstract Information is the wealth of every organization and in the modern-day when information is shared digitally and via the internet, protecting this treasure has become a top issue. Private photographs need to be protected from unwanted access due to security concerns raised by internet photo transfers. Nowadays, practically everyone shares their personal information online, including photographs, either with other users or in a database that attracts cyber criminals who can use it to their benefit. Steganography can be as a security tool to safely transmit secret information because it is one such technique where the presence of a confidential message cannot be detected. This article compares various steganography techniques, including AES, LSB, DCT, DWT, etc. are compared them with each other and also more advanced techniques involving Cryptography and Steganography i.e., sharing secret data using counting-based s and matrix-based to increase security. Keywords Steganography · Cryptography · Least significant bit · Advanced encryption standard · Discrete cosine transform · Discrete wavelet transform · Counting based secret sharing · Matrix based secret sharing

1 Introduction Image processing is translating a physical image to a digital format and performing different operations on it in order to improve the image or extract confidential information that has to be delivered to a particular recipient. It’s a form of distributed signal in which the input is an image, such as a video frame or the picture itself, and the output is another image or image-related features. Graphic processing may also be defined as a method for enhancing raw pictures captured by cameras or sensors on satellites, spacecraft, and aircraft, as well as photos acquired during a range of everyday chores. Over the past five-decade, several approaches have been created in graphic processing. In the rapidly expanding field of computer science, digital R. Tanya Bindu (B) · T. Kavitha AMCEC, Bangalore, Karnataka 560083, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_28

365

366

R. Tanya Bindu and T. Kavitha

imaging, computer processors, and multimedia equipment have all advanced in recent years. Due to their versatility and accessibility, fields that formerly used analog images are now migrating to digital systems. Medicine, video creation, photography, remote sensors, and security monitoring are all good examples. In addition to what would otherwise be studied, these and other sources generate significant volumes of digital photographic material on a daily basis. Digital image processing’s main goal is to extract useful information from photographs. This would be done entirely by computers in an ideal world, with little or no human involvement [1]. It is common knowledge that today, data security has grown to be a top priority. The development of contemporary communication technologies necessitates the use of unique security measures, particularly when it comes to data networks. As the amount of data being transmitted over the Internet grows daily, network security is becoming increasingly crucial. Cryptography and steganography are the two key methods for ensuring security. Both approaches are well-known and often used in information security. Steganography is the practice of concealing information in digital media by embedding secret messages in such a manner that only the sender and the intended recipient(s) can recognize their presence. By using certain cryptographic methods, it is possible to transfer data over the Internet safely while making it impossible for an adversary to intercept or steal private or secret data [2].

2 Application of Image Processing Image processing has become relatively ubiquitous in many different disciplines as vision, image sensors, and computer technologies have evolved. Picture enhancement for better visualization, image compression and transport, and automatic image recognition representation are some of the other applications of digital image processing. The most common uses of digital image processing in military and security applications are low target detection and tracking, arrow direction, vehicle navigation, wide area surveillance, and automatic/aided target detection. One of the goals of image processing in defense and defense systems is to reduce the amount of work that human analysts have to do in order to deal with the growing number of picture data. The process of Steganography System is shown in Fig. 1. The second most challenging aim of image processing is to develop algorithms and methods that will greatly be dealt with the development of autonomous systems which help in decision-making based on all sensory inputs. Recognition system, authorization techniques, multi-dimensional image processing, image processing, video analysis, customizable DSPs for video coding, high-quality color representation, super high resolution image processing, and other applications are examples of digital photo processing applications [3–5].

A Survey on Various Crypto-steganography Techniques for Real-Time …

367

Fig. 1 Steganography system

3 Issues in Image Processing 3.1 Image Compression The use of digital technology in image storage and transfer is a current trend. Creating a digital television transmission necessitates a large amount of bandwidth. In systems such as teleconferencing, a 64-km per the second route is preferred. Even low channel bandwidths (e.g., up to 1 kb per second) are suitable for some applications, such as videophone and video mobile.

3.2 Image Enhancement One of the goals of development is to increase the quality of photos by processing them. The image may be of poor quality due to low contrast, noise, or blurriness, among other factors. Many algorithms have been developed to combat this type of corruption. The most difficult task is removing the degradation without causing signal damage. Sound reduction algorithms, for example, frequently include location, scale, or slide, which, unfortunately, blurs the image’s boundaries. Methods of orientation have been considered, such as slightly sliding along the edges. However, they are most effective when the harm is modest. How to improve the most corrupted photos is a difficult challenge [6].

368

R. Tanya Bindu and T. Kavitha

3.3 Image Recognition The recognition system should, in most situations, categories the unknown input pattern into a set of categories. When the number of classes is small and all members of a class are almost identical, this technique becomes easier. The problem can get much worse if the number of classes is excessive, or if members of the same class have dramatically varied looks. As a result, the most difficult task is learning how to perceive commonplace things. How do you make a system that identifies the word “book,” for example?

3.4 Image Visualization The term “visualization” is frequently used to refer to a component of computer graphics. The major goal is to use three-dimensional object and event models as the foundation for creating visuals or picture sequences. Using sensitive pieces to create dynamic scenes is difficult (like clothes, hair, trees, waves, clouds, etc.). Models must be realistic, yet computational expenses must be affordable [6].

3.5 Image Security Modern photos are sent through the internet to a variety of applications, including police enforcement, military websites, educational, bank data, and private documents. Because these photographs may include sensitive and secret information, their security is of fundamental importance. That means they must be protected while transmitting from attackers or hackers [7]. To put it another way, these photos should be resistant to leaks. There are a number of methods for safe transmitting data which can be used, including picture and data encryption utilizing standard methods.

4 Application of Cryptography in Image Processing An attacker is highly likely to get access to or even attack sensitive real-time visual data transmitted across insecure connections. By encrypting communications to render them unrecognizable, the art of cryptography enables security. By using certain cryptographic techniques, it is possible to communicate data securely over the Internet while making it impossible for an adversary to intercept or steal private or secret data. This is known as cryptography. Encryption and decryption are two fundamental concepts in cryptography; encryption is the transformation of plain text into cypher text, while decryption is the opposite process as shown in Fig. 2. In

A Survey on Various Crypto-steganography Techniques for Real-Time …

369

contrast to cypher text, which is the text that has been encrypted and is ready to be shared, plain text is the text that contains the real message or data that is not encrypted. To encrypt and decode the data, you’ll need a key. Cryptography is being used to turn sensitive information in real-time photographs into incomprehensible data to avoid unauthorized access. Schemes are usually proposed with a high level of security. Slower rates, on the other hand, are always a problem due to their great complexity, which makes them useless in real-time imaging applications. Steganography is the practice of concealing the presence of a communication, whereas cryptography is the practice of concealing the contents of a message. They’re both utilized to keep things safe. However, none of them are capable of meeting basic security requirements such as resilience, undetectability, or capacity. As a result, a new approach called Crypto-Steganography was developed as in Fig. 3, which combines cryptography with steganography to overcome each other’s faults, making it more difficult for attackers to assault or steal critical data [2]. The above-mentioned approaches namely, cryptography and steganography are coupled to give stronger picture protection than either one alone. The use of two layers of security to safeguard embedded information in a combined cryptography and steganography picture security system complicates steg analysis [8].

Fig. 2 Cryptography system

Fig. 3 Crypto-steganography system

370

R. Tanya Bindu and T. Kavitha

5 Literature Survey Data security is a major concern in today’s world that must be handled. Substantially all digital services including internet communication, imaging systems for the military and healthcare, and multimedia systems demand trustworthy security for the transmission and storage of digital pictures. There is a need for security in digital photographs since internet, mobile, and multimedia technology is all advancing more quickly. Therefore, picture encryption methods are required in order to conceal images from such assaults. To conceal images in this system, we employ Advanced Encryption Technique (AES). AES is symmetric block encryption that is designed to replace DES for commercial uses. AES is symmetric block encryption developed to counter DES for commercial reasons. It features a 128-bit block size and a 128, 192, or 256-bit key size [9]. The plain text, which is 128 bits long, is divided into 16 bytes, which are organized in a 4 * 4 matrix with four rows and four columns, each representing a single byte (8 bits) [10]. The AES encryption i.e., cipher key after performing AES algorithm also serves as the key for the image in the cryptosystem. Because AES is extremely secure, the picture under test is also safe. The image-based on AES is also faster than some image cryptosystems based on chaotic systems, according to simulation data [11]. Because it uses distinct keys for encryption and decryption, the RSA algorithm is asymmetric. In terms of encryption and decryption, the RSA method has three main phases. Two keys are created during this stage. The term “public key” refers to a cryptographic key that can be used in public. The receiver can only use this key to decrypt the message [12]. Because of RSA’s asymmetric encryption technique, encryption is safer, and the receiver is less hesitant to give each sender a unique key to ensure communication. Another advantage of the RSA algorithm is that it is difficult to decipher since it includes the factorization of tough-to-factor prime numbers [13]. A statistical metric that illustrates the strength between two variables is the correlation measurement. The correlation coefficient is utilized in this approach to assess how the algorithm and encryption quality are related. The tighter the association and hence the better the image encryption, the closer the coefficient is to zero. In paper specified as [12], the AES algorithm’s correlation coefficient is closer to zero, suggesting higher levels of correlation. The AES technique also provides a higher picture encryption quality since it has more convergent columns in the histogram. Finally, the results of this experiment demonstrated that the AES approach surpasses the RSA algorithm in photo encryption [12]. To prevent assaults, one paper first encrypted a message with the AES algorithm and hashes the key with SHA-2. Following that, they tweaked the LSB method by adding a key to make the hiding process non-sequential [14]. LSBS involves overwriting the bit with the lowest arithmetic value, as the name implies. The end outcome of this method somewhat modifies the original output. The modification is done in such a way that it is unlikely to be detected by human eyes [15]. When using LSB methods to create a grayscale image with a record larger than the message content, each pixel can be encoded with three bits. When it comes to switching over

A Survey on Various Crypto-steganography Techniques for Real-Time …

371

the picture pixels and secret messages, we’ll only encourage the mystery message to be two-fold bits with the two bits of the LSB, we will receive the key. This key will be stego keyed between the transmitter and receiver. Without this stego key, the recipient will be unable to read the secret data. This framework cannot function without this stego key. This key is known as a Dynamic Symmetric key [16]. While this approach works well with 24-bit color picture files, owing to color variations restrictions and the usage of a color map, it hasn’t performed as well with 8-bit color image files. A text message is submerged in the final compressed picture after compression during its implementation. This buried information can only be retrieved by utilizing proper decoding procedures [17]. Another LSB picture steganography technique XOR’s each LSB of the red channel with the secret key to embed the secret data, and then chooses whether to replace the secret data in the green or blue channel. The secret key must then be converted into ten-bit stream arrays. In order to retrieve hidden data, they must split the stego picture into the three matrices RGB. For picture compression, the Discrete Cosine Transform (DCT) is often used. Because of its capacity to compress data with a high level of power, multiplication is an important and basic step in computing the discrete cosine transform (DCT) of a picture [18]. It’s best to use a 2D Discrete Cosine Transform (DCT) with a convenient carry look-ahead adder. By providing a variable approximation at the lowest power and delay in its most accurate mode, the approximate booth multiplier compensates for the bulky hardware required to provide adjustable approximation using individual approximate multiplier modules in 2D-DCT [19]. The Discrete Fourier Transform (DFT) is used in many signal/image processing systems to convert input signals/images from one realm to another. The DFT’s hardware is complicated, and different researchers have proposed several implementation solutions. Distributed arithmetic is one of the most attractive and useful approaches for implementing any discrete orthogonal transform. This research proposes an efficient and effective approach for implementing DFT utilizing distributed arithmetic. In comparison to existing group distributed arithmetic for 8-point DFT, the proposed approach uses a recurring pattern of coefficients, stores effectively in memory, and reduces space by 75% [20]. The Discrete Wavelet Transform (DWT) is a potent technique for obtaining compressed images at greater compression ratios with higher PSNR values. It uses wavelet transforms for image compression. Wavelets handle data discontinuities better than the DCT since it is not Fourier-based, in comparison to the DCT. The DWT is used to illustrate the signal’s dynamic sub-band breakdown. Sub-band analysis is possible without the restriction of dynamic decomposition due to DWT generation in a wavelet packet [21]. The wavelet transform decomposes the signal’s time-domain components using a high pass and low pass filter, resulting in low pass (LL) and high pass (LH) filters for the low pass version and low pass (HL) and high pass (HH) filters for the high pass version [22]. One example of DWT in real-time is the design and chip implementation of 2D-DWT using VHDL programming. Level 1 decomposes the original 2D image (64 × 64) into 32 × 32 size LL, LH, HL, and HH sub-bands. Level 2 decomposes the image into 16 × 16 size sub-bands [23].

372

R. Tanya Bindu and T. Kavitha

The study on Counting Based Secret Sharing (CBSS) created two different modeling variations for the generation of secret shares, namely the 1-bit and 2-bit methods. The CBSS key research explains the trade-offs involved in determining the number of secret shares versus the target key depending on the application’s security requirements. The biggest disadvantage is that as the level of security grows, the proportion of security decreases [24]. They later addressed the flaws in the original CBSS system by proposing a new architecture for secured share distribution that would make pooling and secret restoration easier [25]. Another variant, known as the optimized CBSS, was proposed, which assumed that all phases, including the share generation phase and the share reconstruction phase, were conducted with varied huge sizes of secret keys. In level of security, capacity, and robustness, this experiment used a benchmark of three images of various sizes used to access shares using steganography [26]. We can present a computationally straightforward threshold secret sharing technique in Matrix Based Secret Sharing (MBSS). The method increases the security of already-existing equivalent mechanisms by generating more shares for little secrets at a lower cost than increasing share size [27]. By converting the target key into a matrix, which adds a considerable number of zeros and hence creates a significant number of shares, the matrix- based secret sharing scheme improves upon the secret sharing scheme using counting-based by getting past the zero limitation of the target key [27–30]. To strengthen the security when sending the data, MBSS may be used with any steganography technique like LSB, DWT, etc.

6 Conclusion LSB is one of the above-discussed stenographic methods that are quick and easy, but it has the drawback that if the message is large, it will affect the stego image and the encrypted data can be seen. To overcome the previous drawback Discrete Wavelet Transform can be used. DWT method is simple, fast, and generates the least amount of visibility and picture distortion. Compared to the other strategies mentioned above, DWT approaches are more quickly discovered and offer more security. In terms of quality, DWT has outperformed other steganography methods. The matrix-based secret sharing method using picture steganography, or the DWT, may be encrypted using a variety of encryption algorithms. The stability and security of the system may be enhanced as a result.

References 1. Soni M, Khare A, Jain S (2014) A survey of digital image processing and its problems. 4:2250– 3153 2. Rahmani MKI, Arora K, Pal N (2014) A crypto-steganography: a survey. 5(7)

A Survey on Various Crypto-steganography Techniques for Real-Time …

373

3. Rao RM, Arora MK (2004) Overview of image processing. In: Advanced image processing techniques for remotely sensed hyperspectral data, pp 51–85 4. Billingsley FC (1970) Applications of digital image processing. Appl Opt 9(2):289–299 5. Saxena E, Goswami N (2014) Automatic object detection in image processing: a survey. 2:2321–8169 6. Huang TS, Aizawa K Image processing: some challenging problems. 21:9766–9769 7. Ahmad SS (2016) Steganography for inserting a message on the digital image using least significant bit and AES cryptographic algorithm. 1–6 8. Bhardwaj R, Khanna D (2015) Enhanced the security of image steganography through image encryption. 1–4 9. Deshmukh P (2016) An image encryption and decryption using AES algorithm. 7:2229–5518 10. Aparna VS, Rajana A, Jairaja I, Nandita B, Madhusoodanana P, Remya AAS (2020) Implementation of AES algorithm on text and image using MATLAB. 978-1-7281-4213-5 11. Zhang Y, Li X, Hou W (2017) A fast image encryption scheme based on AES. 978-1-50906238-6 12. Alsaffar DM, Sultan Almutiri A, Alqahtani B, Alamri RM, Fahhad Alqahtani H, Alqahtani NN, Mohammed alshammari G, Ali AA (2019) Image encryption based on AES and RSA algorithms. 978-1-5386-9439-8 13. Jain A, Sharma S (2019) A novel digital image encryption method based on RSA algorithm. 11. ISSN: 0973-7383 14. AL-Shaaby AA, AlKharobi T (2017) Cryptography and steganography: new approach. 5. ISSN: 2054-7420 15. Siper A, Farley R, LomBardo C (2005) The rise of steganography. Corpus ID: 110715828 16. Amarendra K, Mandhala VN, Gupta BC, Sudheshna GG, Anusha VV (2019) Image steganography using LSB. 8. ISSN: 2277-8616 17. Aggarwal A, Sangal A, Varshney A (2019) Image steganography using LSB algorithm. 11. ISSN: 0974-2255 18. Zhou Y, Wang C, Zhou X (2018) DCT-based color image compression algorithm using an efficient lossless encoder. ISSN: 2164-5221 19. Kumar YS, Kumar R, Kumar S (2020) 2D-Discrete cosine transform based dynamically controllable image compression technique. 978-1-7281-8911-6 20. Kumar SRS, Veeramachaneni S, Sk NM (2019) An efficient DFT implementation using modified group distributed arithmetic. 978-1-7281-1381-4 21. Kour P (2015) Image processing using discrete wavelet transform. 3. ISSN: 2321-5984 22. Thakral S, Manhas P (2019) Image processing by using different types of discrete wavelet transform. 499–507 23. Bisht A, Kumar A (2019) DWT chip design and FPGA synthesis for image processing. 8. ISSN: 2277-3878 24. Gutub A, Al-Juaid N, Khan E (2017) Counting-based secret sharing technique for multimedia applications. 5591–5619 25. Gutub A, Al-Ghamdi M (2019) Image based steganography to facilitate improving countingbased secret sharing. https://doi.org/10.1007/s13319-019-0216-0 26. Gutub A, Al-Ghamdi M (2020) Hiding shares by multimedia image steganography for optimized counting-based secret sharing. 7951–7985 27. Porwal S, Mittal S (2020) A threshold secret sharing technique based on matrix manipulation. 2214:020020 28. Al-Shaarani F, Gutub A (2021) Securing matrix counting-based secret-sharing involving crypto-steganography. https://doi.org/10.1016/j.jksuci.2021.09.009 29. Vinothkanna MR (2019) A secure steganography creation algorithm for multiple file formats. J Innovative Image Process (JIIP) 1(01):20–30 30. Chatterjee R, Chakraborty R, Mondal JK (2019) Design of lightweight cryptographic model for end-to-end encryption in IoT domain. IRO J Sustain Wireless Syst 1(4):215–224

A Lightweight Image Cryptosystem for Multimedia Internet of Things V. Panchami, Arjun Rajasekharan, and Mahima Mary Mathews

Abstract The popularity of social media websites tends to boost the volume of multimedia material. This results in the development of the brand-new industry known as the Multimedia Internet of Things. Lightweight image encryption methods are required to protect multimedia data on these resource-constrained devices. Since chaos theory has grown more common in modern multimedia cryptography, it serves as the foundation for the suggested lightweight encryption technique. 2D augmentation models are used in this chaotic-based multimedia encryption method to provide secure data transit. The suggested method has minimum residual clarity and key sensitivity while, simultaneously maintaining the excellent encryption quality of chaotic maps. The simplified image encryption technique employs the chaotic map model, which has the properties of confusion and diffusion. We have also put forth a novel key generation algorithm for use with the Logistic map and the Rubik’s cube transformation. Using the Elliptic-Curve Cryptography (ECC) Key Algorithm, the initial values are produced. To analyze the computational complexity, the suggested method is applied to medical (binary) and colored images. The histograms of the encrypted images are flat and distributed across all the pixel values, according to the security analysis. These images have an entropy of 7.86046675 with average correlation values of 0.0010575 (horizontal), 0.013994 (vertical), and 0.00235 (diagonal) (encrypted image). The suggested lightweight image encryption hence displays a high level of security.

V. Panchami (B) CSE-Cyber Security Department, Indian Institute of Information Technology, Kottayam, Kerala, India e-mail: [email protected] A. Rajasekharan Department of Computer Science and Engineering, Design and Manufacturing, Indian Institute of Information Technology, Kancheepuram, Tamil Nadu, India M. M. Mathews Department of Computer Science and Engineering, Indian Institute of Information Technology, Kottayam, Kerala, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_29

375

376

V. Panchami et al.

Keywords Multimedia Internet of things · Image encryption · Chaotic maps · Elliptic-curve cryptography key algorithm · Arnold’s cat map · Henon map · Logistic map

1 Introduction The tremendous growth in the Internet of Things and multimedia data arise a situation that multimedia data is particularly susceptible, and its protection is more important. Multimedia data is being disseminated at an alarming pace across physical media and the Internet, making data security a critical issue. Data security may prevent data breaches and ensure privacy by encrypting the audiovisual aspect. Therefore, lightweight approaches have to be adapted to secure multimedia data which offers low computational complexity. The conventional encryption techniques cannot be applied to secure communications in the Multimedia Internet of things as it has high computation cost and complexity. Moreover, Asymmetric encryption techniques have a higher computational cost than symmetric encryption algorithms, in which the image is represented by a vector of real numbers. As a result, the vector is rather lengthy due to the image’s enormous sampling coefficients. In computing, a digital image is a 2D vector with pixels ranging in value from zero to two hundred fifty-five. These numbers can be used to represent any geometric shape (circles, curves, or lines) in an image. As a result, safeguarding digital images is critical. Furthermore, the field of information technology is continuously evolving. As a result, cloud computing has grown widely across many industries, including manufacturing, railways, commerce, and government, where security is paramount. As a result, the security architecture is required in these industries to prevent unauthorised individuals from accessing the cloud. Surveillance system applications have developed dramatically during the previous two decades. These systems may be found all over the world (public or private). The cloud, surveillance systems, railways, specialised networks, and applications in the medical field may all benefit from the lightweight image encryption technique. A highly effective image encryption approach is critical for safeguarding content from illegal access and dissemination. Data privacy and security are significant concerns in today’s Internet world. By virtue of its low computational cost, the lightweight image encryption approach creates an intriguing framework. This study makes the following significant contributions: • This approach is developed in order to ensure that neighbouring pixels of the encrypted picture have lower correlation coefficients and requires minimal time and utility for key modification. • To encrypt and decrypt the images using a highly secure cipher. • To design a novel Key generation algorithm based on Rubik’s cube transformation and Logistic Map. • To analyse the security of the proposed system.

A Lightweight Image Cryptosystem for Multimedia Internet of Things

377

2 Chaos-Based Encryption and Decryption 2.1 Chaos Maps Chaotic systems are a subset of nonlinear dynamical systems [1] that are easy to comprehend. They may have a small number of interacting elements and operate according to the basic principles, yet all of these systems exhibit a strong sensitivity to their baseline conditions [2]. In the context of chaos, a map (or function) is an equation or a rule that describes the evolution of a dynamic system through time.

2.2 Chaos Maps for Encryption Traditional encrypting algorithms such as AES and RSA have many disadvantages and shortcomings [2] when it comes to the encryption of digital images and high processing power: • Excessive computational time for big images. • Excessive computational power for large images. As a result, there may be more advanced approaches for image encryption. A few chaos-based algorithms provide an excellent trade-off between speed, security complexity, and computational overhead.

2.3 Elliptic-Curve Cryptography Key Algorithm The ECDH (Elliptic Curve Diffie-Hellman Key Exchange) is a mutual authentication system that enables the sender and receiver will generate a secret key on both sides while utilizing the public–private key pairs. ECDH is quite similar to Diffie-Hellman Key Exchange, except that it utilises ECC point multiplication instead of modular exponentiations. The ECDH is centered on this property: ( p ∗ G) ∗ q = (q ∗ G) ∗ p Consider: There may be exchange of the secret number ( p ∗ G) and (q ∗ G) (the public keys of P1 and P2) across an unsecured channel and derive a shared secret if there are two secret numbers p and q (two private keys belonging to P1 and P2) and an ECC elliptic curve with generator point G.

378

V. Panchami et al.

3 Proposed Lightweight Image Encryption Algorithm The proposed lightweight image encryption algorithm uses Henon Map, XOR, to achieve diffusion property and Arnold Cat Map (ACM) [3, 4] for confusion property. These operations have two inputs, the image which is represented as the M × M matrix and the keys K 1, K 2 and K 3. The initial values of the keys are generated using Elliptic Curve Cryptography (ECC) key generation algorithm by utilizing the private key of the sender and the public key of the receiver which will be 32 values as in Fig. 3. These 32 values are compressed into 24 values using a compression function while applying the XOR operation. The 24 values are then represented as a (2 × 2 × 2) three dimensional matrix for the generation of K 1, K 2 and K 3. The 3-D key generation matrix will be undergone through a transposition operation based on Rubik’s Cube. The output of this transformation will be treated as the first key K 1. The key K 2 will then act as an input to the logistic Map to produce the second key K 2. The third key will be the result of the XOR operation between the first key K 1 and second key K 2. Figure 1 depicts the overall architecture of the system.

3.1 Key Generation Algorithm The ECC based key generation algorithm will generate the initial key IK which has 32 values (32 × 4 = 128 bits). This 32 values will be input to a compression function using XOR which is will be represented as follows: I K i ← I K i ⊕ I K i+8 The output from this function will be 24 values. As the algorithm show below, this 24 values are represented in 3D key generation matrix. Chaotic systems are a subset of nonlinear dynamical systems [1] that are easy to comprehend. They may have a small number of interacting elements and operate according to extremely basic principles, yet all of these systems exhibit a strong sensitivity to their baseline conditions [2]. In the context of chaos, a map (or function) is an equation or a rule that describes the evolution of a dynamic system through time. Notations and Explanation are shown in Table 1. 3D Generation Matrix and Rubik’s Cube Transformation The 3D key generation matrix has 6 sides, F (face) which is represented as RED, R (right) as ORANGE, L (right) as BLUE, U (up) as YELLOW, B (bottom) as GREEN and B (bottom) as white. Each face is again divide into 4 equal parts TL (Top left), TR (Top right), BL (Bottom left) and BR (Bottom right). The Rubik’s cube transformation is done by applying six conditions for each side of the cube. For example, if the number is “zero” in the Top left, TL position of the face F of the cube (red, R) then the Matrix should be rotated in a Zig Zag form as: “R2 F2 R2 U2” otherwise it should be rotated

A Lightweight Image Cryptosystem for Multimedia Internet of Things

Fig. 1 Proposed chaotic based image encryption scheme Table 1 Notations and explanation Notation

Explanation

R

R Orange face of the cube

F

F Red face of the cube

U

U Yellow face of the cube

D

D White face of the cube

L

L Blue face of the cube

B

B Green face of the cube

Mx

Mx Matrix which consists of elements of face X

X

X Face X of the cube will be rotated in clockwise direction

X

X Face X of the cube will be rotated in anti-clockwise direction

Xn

Xn Face X of the cube will be rotated “n times” in clockwise direction

Xn

Xn Face X of the cube will be rotated “n times” in anti-clockwise direction

379

380

V. Panchami et al.

as Quarter’s form, “U F2 U2 R2 U”. The result of the final Matrixes are concatenated with the last 6 values of the initial key, to form the first key K1 having 30 values. Logistic Map based Second Key Generation The second key is generated using Logistic Map. The first key K1 will be the input of the Logistic Map. Each values in K1 is treated as Blocks B1 , B2 . . . to Bn . Block means the binary values of each digit. X0 will the initial condition from Logistic map. X01 is the sum of first 3 Blocks B1, B2 and B3 then divide it by 224 . The next value X01 is the sum of next 6 Blocks, B4, B5, B6, B7, B8 and B9 which is divided by 16 × 6. Then calculate X 0 ← X 01 + X 02 . For the next cycle X01 is generated as X 01 ← B Pr evious cycle + 3 to B Pr evious cycle and X02 as X 02 ← B Pr evious cycle + 6 to B Pr evious cycle . The output as the Logistics Map operation is treated as the second key K2. The XOR operation is applied between the first key K1 (30 values) and the second key K2 (30 values) to generate the third key K3.

3.2 Algorithm for Key Generation Input: Initial Key (IK: 32 digits) Output: Keys K1, K2, K3 1. 2. 2.1. 2.2. 2.3. 3. 3.1. 3.2. 3.3. 4. 4.1. 4.2. 4.3. 5. 5.1. 5.2. 5.3. 6. 6.1. 6.2. 6.3. 7. 7.1.

i←1 if (i ≤ 26) I K i ← I K i ⊕ I K i+8 Print I K i and store the value to M, 3D matrix (2 × 2 × 2). Increment i (i++) if (M R (TL) == 0) then M R ← Zig − Z ag: R2 F2 U 2 else M R ← Quar ter s: U F2 U 2 R2 U if (Mo (BL) == 1) then Mo ← (R  D  R) ∗ 4U  (R  D  R D) ∗ 2 Else Mo ← (R  D  R) ∗ 2U  (R  D  R D) ∗ 4 if (M B (TR) == 0) then M B ← Cubic: R F U  R2 U F  R U F2 R2 else M B ← Square: U R F2 U R F2 R U F  R if (MG (BR) == EVEN NUMBER) then MG ← Pillar: U R U  R2 U  R  F  U F2 R F  else MG ← Spiral: U 2 F2 R2 U 2 R if (MW (TL) == ODDUMBER) then MW ← R D B2D  R 

A Lightweight Image Cryptosystem for Multimedia Internet of Things

7.2. 7.3. 8. 8.1. 8.2. 8.3. 9. 10. 10.1. 10.2. 10.3. 10.4. 10.5. 10.6. 10.7. 10.8. 10.9. 11. 12.

381

else MW ← D  R  D if (MY (BL) == EVEN NUMBER) then MY ← U RU  L  U R  U  L else MY ← F L F L  D  L  D First Key, K 1 ← M R v Mo v M B v MG v Mw v MY v Last 6 digits Compute Second Key, K 2 ← LogisticMap(K 1 ) X o ← Initial condition from LogisticMap Bi ← Blocks from First Key X i ← X j , wher e j = 1, 2 X 01 ← (B1 + B2 + B3 )/224 X 02 ← (B4 + B5 + B6 + B7 + B8 + B9 )/16 × 16 X 0 ← X 01 + X 02 For the next cycle X 01 ← B Pr evious cycle + 3 to B Pr evious cycle + 5 X 02 ← B Pr evious cycle + 6 to B Pr evious cycle + 5 Third Key, K 3 ← K 1 ⊕ K 2 Print K 1 , K 2 , K 3 .

3.3 Encryption Process The input of the lightweight image encryption is the original image. This original image will undergone through a reshaping mechanism in which it composed of two main stages: preparation and pixel modification. The image dimension is rectangular and the image colour model is RGBA (red, green, blue, alpha). For instance, a 400 × 600 image becomes a 600 × 600 image. It will be updated with new pixels with an alpha coefficient of 254, rather than 255. A pixel’s transparency is controlled by the RGBA colour model’s Alpha channel. To distinguish them from the original image pixels, the subsequently inserted pixels are assigned the value of 254. The new pixels range from 0 to 255. Pixel manipulation has two stages: confusion and diffusion. Both stages need beginning values. These results are calculated using the keys K1, K2 and K3.

3.4 Confusion Process Utilising ACM This image encryption system’s first-pixel alteration is called confusion [5]. The picture is shuffled or randomized such that it can no longer be recognised. In this architecture, ACM [6] is employed to calculate pixel displacement. ACM [7] needs three positive parameters: p, q, and the amount of repetitions. Its values originate from the hidden key element values. The ACM formula is [8]:

382

V. Panchami et al.



xn yn 





1 p = q pq + 1



xn yn

 mod N

The above formula yields a matrix with xn and yn as the new position. The pixels in position (xn , yn ) will be relocated to (xn  , yn  ). This computation is repeated until all pixels are shuffled. The mesh grid function of Python NumPy [9] may be used to efficiently move pixel coordinates in a matrix. First, create a N × N mesh grid for x and y to represent the original pixel location in the picture. Two mapping matrices are generated for the new pixel position using the aforementioned algorithm. That is, the [x, y] pixel value shifts according to the matrix. This process is performed for the stated number of times. To make the output matrix less predictable [10], the values of p and q are varied by two integers from the hidden key.

3.5 Diffusion Process Utilising Henon Map This diffusion technique works by employing bitwise OR gate logical operation on a matrix of image pixels created by a chaotic map matrices [11]. The matrix is Henon Map. The matrix requires two starting variables, xn and yn . Both starting values come from hidden key values, much as the confusion stage. The values of xn and yn are 0–1 in this scheme. This implies that slight changes [12] in values may have big effects on the diffusion. The chaotic map is built by bit-wise XORing the image pixels produced by the confusion phase [13]. A new vector is created using the diffusion findings. In order to remove the high correlation between neighboring pixels, the zigzag approach was employed to scramble the pixel positions of the image [14]. The process began with the first pixel of the image matrix, and then proceeded to traverse the consecutive pixels in two dimensional zigzag mode, converting the two-dimensional matrix into a one-dimensional sequence. The starting position for a matrix of size 3 × 3, for example, begins at element 1. Figure 2 shows the original matrix and one-dimensional sequence after a 2D zigzag scan.

3.6 Decryption Process Decryption is the inverse process of encryption, decryption includes inverse of diffusion and confusion as well as crop border processes. The altered image pixels are used to restore their native values and placements. Both phases need initial values. The hidden key is used to get these attributes’ data. To decode the image back to its original form, the inverse of diffusion and confusion [15] is done with the starting value parameters which should be matched with those used in the encryption procedure. Value differences may prevent decryption. After completing the inverse of diffusion and confusion steps, the crop border procedure must be finished. The original image may be seen at this phase, however the image proportions have changed owing to

A Lightweight Image Cryptosystem for Multimedia Internet of Things

383

Fig. 2 3-dimensional key generation matrix

the decryption process. So, to restore the image’s original proportions, a crop border is used to eliminate the pixels created during the decryption process. The original image may be viewed here, however the decryption process has modified the image dimensions. Then this operation is utilised to remove the pixels formed after the decryption process. Original, Encrypted and Decrypted Images- RGB are shown in Fig. 4. Original, Encrypted and Decrypted Images (Binary) as shown in Fig. 5.

4 Results and Discussions The security analysis performed in this paper are Intensity Histogram analysis, Adjacent Pixel Auto Correlation analysis and computational complexity is evaluated to evaluate the performance.

4.1 Input and Output Cases The experiment has been performed in Python 3.9.0 using Visual Studio Code, an environment with a computer of 6 core i7, 2.2 GHz and 32 GB RAM. This test included three RGB images (512 × 512), three medical (binary) images (256 × 256), and a grayscale image (256 × 256).

384

Fig. 3 Key generation algorithm

V. Panchami et al.

A Lightweight Image Cryptosystem for Multimedia Internet of Things

385

Fig. 4 Original, encrypted and decrypted images (RGB)

4.2 Intensity Histogram Analysis Image histogram analysis is a simple way to show image encryption quality. Image encryption converts plain text images into random, unintelligible forms. A decent image encryption method produces a cipher image with a consistent intensity histogram. Figures 6 and 7 show the histogram analysis of the coloured and grayscale images. The uniform histogram indicates the best effective encryption technique.

386

V. Panchami et al.

Fig. 5 Original, encrypted and decrypted images (binary)

4.3 Entropy Analysis Information entropy gauges an encryption system’s unpredictability. a typical encryption algorithm’s major criteria More entropy usually means a stronger encryption technique. The tables present statistical data for several image groups of varying sizes. For analysis we have considered the encrypted image’s entropy, NPCR (number of changing pixels per second), UACI (unified averaged changed intensity), PSNR (peak signal-to-noise ratio), and MSE (mean squared error). Observations from Tables 2 and 3: i. As the image size grows, the NPCR, UACI, and entropy values grow as well. This is an efficient image encryption method. ii. General (mixed) images have the least entropy.

A Lightweight Image Cryptosystem for Multimedia Internet of Things

387

Fig. 6 Histogram of Lena (RGB)

4.4 Computational Complexity Analysis Various image groupings were studied for computational complexity [16] in terms of time. We need an optimization technique to find the optimal beginning value. The encryption and decryption times are image-independent. However, these time frames vary with image size. The computational performance of encryption and decryption are similar. The dimensions of the image has the highest impact over time complexity [17]. As per the analysis results given in Tables 4 and 5, the proposed system has very less computation complexity for coloured images and it is very efficient for grayscale images.

4.5 Adjacent Pixel Auto Correlation Analysis The correlation of neighbouring pixels [18, 19] are utilized to measure encryption performance (Horizontal, Vertical or Diagonal). The correlation coefficients [20] of encrypted images are clearly near to zero, indicating that our suggested approach can efficiently eliminate correlations between neighbouring pixels and withstand statistical intrusions [21]. Correlation coefficients between adjacent pixels for different images on RGB and Binary are shown in Tables 6 and 7.

388

V. Panchami et al.

Fig. 7 Histogram of medical image (binary)

Table 2 Statistical analysis for different groups of images (RGB) Image (512 × 512) NPCR (%) UACI (%)

Entropy (plain image) Entropy (cipher image)

Lena

99.636296

23.517091 7.271856

5.988000

Underwater image

99.962997

34.376602 2.188325

5.736782

Texture (grass)

99.60289

23.826393 6.809031

5.980303

Table 3 Statistical analysis for different groups of medical images (binary) Medical image (256 × 256)

NPCR (%)

UACI (%)

MSE

PSNR (dB)

Entropy (plain Entropy image) (cipher image)

MI_1

99.858856

48.823907

21,062.81

7.89564

1.397531

MI_2

99.818039

48.911065

21,117.63

7.88435

1.884436

7.997107

MI_3

99.821472

48.097165

20,596.95

7.992773

2.532671

7.997207

7.997295

A Lightweight Image Cryptosystem for Multimedia Internet of Things

389

Table 4 Analysis of obtained computational complexity in terms of time for different images (RGB) Image (512 × 512)

Size

Encryption (s)

Decryption (s)

Total time (s)

Lena

512 × 512

1.9602

2.0887

4.0489

Underwater image

512 × 512

1.9255

2.0951

4.0206

Texture (grass)

512 × 512

1.9119

1.9998

3.9117

Table 5 Analysis of obtained computational complexity in terms of time for different medical images (binary) Medical image (256 × 256)

Size

Encryption (s)

Decryption (s)

Total time (s)

MI_1

256 × 256

0.4694

0.4952

0.9646

MI_2

256 × 256

0.5183

0.6097

1.1280

MI_3

256 × 256

0.5158

0.5538

1.0696

Table 6 Correlation coefficients between adjacent pixels for different images (RGB) Image (512 × 512)

Plain image Horizontal

Vertical

Diagonal

Horizontal

Vertical

Diagonal

Lena

0.979578

0.988129

0.972011

−0.00011

0.0024

−0.0012

Underwater image

0.942922

0.943511

0.901062

−0.002

−0.0025

−0.0017

Texture (grass) 0.781247

0.800600

0.726798

−0.0019

7.2259 × 10−4

−0.0021

Cipher image

Table 7 Correlation coefficients between adjacent pixels for different groups of medical images (binary) Medical image (256 × Plain image 256) Horizontal Vertical

Cipher image Diagonal Horizontal Vertical

Diagonal

MI_1

0.959446

0.940174 0.913983 0.007529

0.002932 −0.005537

MI_2

0.954145

0.934506 0.909612 0.008509

0.002016 −0.001271

MI_3

0.923748

0.887473 0.846889 0.007303

0.002962 0.000146

5 Conclusions The results of this study lead to many conclusions in which the proposed Image encryption algorithm is very efficient for both grayscale images and coloured images. The histogram distribution of cipher images are uniform. Moreover, using the ACM in the confusion phase with varying p and q values renders pixel placements unpredictable and difficult to anticipate. The starting values of the Henon map are at least 1014 sensitive. Therefore, changing the beginning settings of the ACM, Logistic, or Henon map may significantly alter the encryption and Key generation results. A 128

390

V. Panchami et al.

bit numeric key provides great security since it can produce 2128 variations. Therefore it has high resistance against brute force attack. The biggest dimension of an image (length/width) has the greatest impact on performance and time. This technique uses a single confusion layer and numerous diffusion layers. Chosen chaotic map is exhibits complex non-linearity and choosing a high-dimensional chaotic system expands the key space.. The diffusion function changes pixel values but does not repeat permutations. The new technique has been subjected to several security evaluations, and the results demonstrate that the proposed algorithm exhibits very strong cryptographic properties, less computation, efficient and adaptability to implement in lightweight applications such IoT devices.

References 1. Desai AP, Crasto J (2012) Chaos-based system for image encryption, pp 4809–4811 2. Fadhel S, Shafry M, Farook O (2017) Chaos image encryption methods: a survey study. Bull Electr Eng Inf 6(1):99–104 3. Soleymani A, Nordin MJ, Sundararajan E (2014) A chaotic cryptosystem for images based on henon and Arnold cat map. Sci World J 4. Kocarev L (2001) Chaos-based cryptography: a brief overview. IEEE Circuits Syst Mag 1(3):6– 21 5. Kumar Y, Mamta S (2014) A review paper on image encryption techniques. Int J Res Appl Sci Eng Technol 5(4):169–172 6. Abbas NA (2016) Image encryption based on independent component analysis and Arnold’s cat map. Egypt Inf J 17(1):139–146 7. Hariyanto E, Rahim R (2016) Arnold’s cat map algorithm in digital image encryption. Int J Sci Res (IJSR) 5(10):1363–1365 8. Elzaher MFA, Shalaby M, Ramly SHE (2016) An Arnold cat map-based chaotic approach for securing voice communication. In: The 10th international conference on informatics and systems, Giza, pp 329–331 9. Elshamy AM, Abdelghany MA, Alhamad AQ, Hamed HFA, Kelash HM, Hussein AI (2017) Secure voip system based on biometric voice authentication and nested digital cryptosystem using chaotic Baker’s map and Arnold’s cat map encryption. In: 2017 international conference on computer and applications. ICCA 2017, Doha, pp 140–146, 2017.8079739 10. Saha C, Hossain MF (2019) Mri watermarking technique using chaotic maps, NSCT and DCT. In: 2nd international conference on electrical, computer and communication engineering. ECCE 2019, IEEE, Cox’s Bazar 11. Dyson FJ, Falk H (1992) Period of a discrete cat mapping. Am Math Mon 99(7):603 12. Munir R (2019) A secure fragile video watermarking algorithm for content authentication based on Arnold cat map. In: 2019 4th international conference on information technology (InCIT), IEEE, Bangkok, pp 32–37 (2019) 13. Raghava NS, Kumar A (2013) Image encryption using henon chaotic map with byte sequence. Int J Comput Sci Eng 3(5):11–18 14. Lin J, Si X (2009) Image encryption algorithm based on hyperchaotic system, pp 153–156 15. Afifi A (2019) A chaotic confusion-diffusion image encryption based on Henon map. Int J Netw Secur Appl 11(4):19–30 16. Sekar JG, Arun C (2020) Comparative performance analysis of chaos based image encryption techniques. J Crit Rev 7(9):1138–1143 17. Kamali SH, Shakerian R, Hedayati M, Rahmani M (2010) A new modified version of advanced encryption standard based algorithm for image encryption. In: Kyoto J (ed) Proceedings of the international conference on electronics and information engineering, pp 141–145

A Lightweight Image Cryptosystem for Multimedia Internet of Things

391

18. Zhou Y, Panetta K, Agaian S (2009) Image encryption using binary key images. In: Proceedings of the IEEE international conference on systems, man and cybernetics. Hyatt Regency Riverwalk, San Antonio, TX, USA, pp 4569–4574 19. Zhang X, Han F, Niu Y (2018) Chaotic image encryption algorithm based on bit permutation and dynamic DNA encoding. Comput Intell Neuro Sci 11(2017):1–12 20. Bisht A, Dua M, Dua S, Jaroli P (2019) A color image encryption technique based on bit-level permutation and alternate logistic maps. J Intell Syst 342:1–15 21. Chatterjee R, Chakraborty R, Mondal JK (2019) Design of lightweight cryptographic model for end-to-end encryption in IoT domain. IRO J Sustain Wirel Syst 1(4):215–224

A Study on Parking Space Allocation and Road Edge Detection for Optimizing Road Traffic H. Varun Chand, Seema Sabharwal, Anil Carie, and S. Arun Kumar

Abstract Road traffic has been increasing rapidly for the past few decades, which has resulted in air pollution, long waiting times in road traffic, frustration, wastage of effective time on the road etc. The driver’s poor vision in identifying road edges, especially during the night, may result in major or minor accidents, which all lead to traffic congestion. Finding a parking space during peak hours also results in congestion. With the emergence of machine learning, IoT, computer vision etc., these problems can be minimised. Road edge detection helps to prevent vehicles that are running off the road and parking space allocation helps to allocate unoccupied space for vehicles even at peak times. The paper addressed a study on different methods through which road edge detection and parking space allocation is possible. This literature survey helps researchers to understand the importance of various cutting edge techniques in road edge detection and parking space allocation. Based on the study, a model is being proposed to provide a solution to edge detection of roads and allocate parking slots, thereby optimising the road traffic. Keywords Vehicular ad-hoc network · Road edge detection · Parking space allocation · Computer vision · Machine learning

H. Varun Chand (B) Department of Computer Science and Engineering, College of Engineering Perumon, Kollam, Kerala, India e-mail: [email protected] S. Sabharwal Department of Computer Science, Government P.G. College for Women, Panchkula, Haryana, India A. Carie School of Computer Science, VIT-AP University, Amaravathi, India e-mail: [email protected] S. Arun Kumar Department of Computer Science and Engineering, Bethesda Institute of Technology and Science, Gwalior, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_30

393

394

H. Varun Chand et al.

1 Introduction Nowadays, everyone has their own vehicles and it is found to be difficult to travel along the road during rush hours [1]. Due to the massive growth in vehicle numbers, it becomes very difficult to park vehicles in the city premises during peak hours. This resulted in driver frustration, wastage of time, air pollution etc. Also, while traveling along roads in unfavourable circumstances like foggy, rainy etc., it becomes more difficult for the drivers to identify the edges of roads, which results in driver frustration [2]. If the driver cannot identify the edge of the road, it may sometimes result in accidents. During rush hours, it becomes more difficult to park a vehicle even if some vacant parking slots are available. This is due to the lack of proper knowledge about that parking space [3]. So, identifying all available vacant slots and informing the drivers about those may help to solve this problem. But unfortunately, so many factors need to be considered if a registered system allocates parking space, like if two or more drivers are allocated the same vacant slot to whom that space needs to be allocated, if the vacant space is allocated by a non-registered user, etc. Even the size of a vehicle also needs to be considered before allocating the space and all these make parking space allocation a challenging task [4, 5]. The road edge detection will identify off road regions and on road regions. During unfavourable driving conditions like heavy rain, fog, night time etc., the driver feels difficulty in identifying the road edge and sometimes may result in accidents. The second largest number of counts for road accidents in the world is caused by road departure accidents, which is due to lack of identifying the road edge in unfavourable weather conditions [6]. This paper included a review on such techniques through which road edges can be detected. These two problems have been discussed in this paper and provided a detailed literature review on various image processing, computer vision, and machine learning techniques to solve such problems.

2 Study on Parking Space Allocation Due to lack of proper knowledge about vacant lots, drivers find it difficult to allocate vehicles during rush hours. Hence, this results in air pollution, frustration, excess fuel consumption, wastage of effective time etc. Context: A parking space reservation system that can allocate space based on auction. Here the V2G concept is used with the parked vehicle to sell their electricity through electric discharge devices [7]. Objective: An auction based reservation system for parking space allocation with focus on V2G technology is proposed. Method: The user can determine the parking place and time in advance, thereby reducing the waiting time for parking. Here, V2G technology is also used for the

A Study on Parking Space Allocation and Road Edge Detection …

395

storage of electricity for electric vehicles. The parking systems are not based on traditional methods, first come first serve, but based on the willingness of users who pay more money. Hence, it’s called an auction-based system. The system invites reservations from users with a deadline, including their willingness to pay per unit of time, scheduled time of arrival, and scheduled departure time. Some parking lots are equipped with electric discharge devices, where there is a facility for electric vehicles to sell their power. The allocation will be made at the scheduled deadline. Result: The method focused on auction-based systems where the user who pays more money will be allocated the space. Since the space allocated has the facility for electric discharging devices, each of these parking lots containing electric vehicles may act as a power generating source. Significance: V2G is an emerging technology that focuses on storing electricity from electric vehicles, which has been incorporated into allocating parking spaces. Drawback: Since V2G technology is used, if any non-electric vehicles are allocated to parking lots with electric discharge devices, the intended benefit may not be received. Context: CERP-IoT, a project proposed by the European Union to promote various technologies ITS, IoT, etc. This cloud based system can be used for smart parking mechanisms. Objective: A cloud based mobile application for parking systems is implemented with three layers- sensor layer, communication layer and application layer [8]. Method: This cloud based system used three layers. The sensor layer deals with technologies like RFID, infrared, microwave, CCTV etc. for vehicle parking access control. The communication layer provides connection between the sensor layer and the application layer. It uses 3G, 4G, ZigBee, WiFi, V2X etc. for communication. The application layer uses the cloud tier and mobile app tier. Mobile apps are used by the registered drivers and data processing and optimisation is done at the cloud tier. With the help of sensor layer technologies, vacant and occupied parking lots are identified. The cloud platform will process the space allocation and is informed through the registered app. Result: A cloud-based application for parking space allocation that processes all information in three layers. Information is stored and processed at the centralised cloud storage unit. Significance: All data has been processed on the cloud platform, from where the vacant space is allocated for the requested vehicle. Context: A cloud-based smart vehicle parking system (SVPS) is used to allocate parking space for requested vehicles [9]. Objective: Cloud infrastructure is utilised for allocating vacant parking slots for the requested driver and hence overcomes the problem of parking space allocation on time.

396

H. Varun Chand et al.

Method: SVPS architecture has a cloud based parking slot management system and traffic management bureau. This bureau is connected by communication terminals, roadside units, parking side units and vehicular nodes. Here, each of these parking slot units is connected with the roadside unit, which is linked with a cloud server. Each PSU collects information about occupied and vacant slots, which is updated on the cloud server. If a specific user is requesting parking based on availability, the slots will be allotted. If a specific parking lot is not available for booking, the system will find and recommend the nearest lot for parking. The architecture of cloud based SVPS is shown in Fig. 1. Result: It minimises parking search time and improves the performance of parking resource utilisation, thereby reducing fuel consumption. Drawback: It does not provide privacy and security for the data during vehicular communication. The data is prone to attacks. Context: A micro-controlled based system that uses image processing techniques to identify parking locations for different vehicles [10]. Objective: To allocate parking areas using a smart system based on microcontroller and image processing methods for registered users.

Fig. 1 The architecture of the cloud-based SVPS using VANET

A Study on Parking Space Allocation and Road Edge Detection …

397

Method: Optical Recognition technique is used to recognise whether a user is registered or not. The number plate of the vehicle is scanned and from this, the driver details are fetched and are sent to the server. The server processes this data and based on available parking slots, the location route is sent to the registered user through the user app. This system works on a specified private campus area. Result: It allocates spaces that are near to the user. If no such spaces are available nearby, it will allocate space outside the campus area. Drawback: It does not specify unregistered users. If such users allocate parking space, it creates problems and may degrade performance. Context: A real-time monitoring system that collects vacant spaces and, based on requests received from drivers, the vacant space is allocated for parking. It makes use of feature extraction and pattern recognition techniques [11]. Objective: To overcome parking issues, a three-fold method is used for parking space allocation. It involves image collection, preprocessing these images using feature extraction and Bayes classifier techniques, and finally utilising the space using an app. Method: It’s a three-step process. Parking space related information is collected from the driving onboard cameras of vehicles, which are preprocessed and used for detection of vacant parking slots. The parking lots are equipped with various IR sensors, the information collected is transferred to Node MCU which will be classified as allocated and unallocated parking space details to the firebase. The firebase also contains the in-time and out-time of each vehicle. This information is later used for allocation of parking space to a requested vehicle. The architecture is shown in Fig. 2. The paper used feature extraction and pattern recognition for preprocessing the collected images. Result: The study results show this method has a precision of 0.9967and recall of 0.9967.

3 Study on Road Edge Detection in Unfavourable Weather Conditions A large number of road accidents occur due to vehicle off-road collisions. The drivers are unable to detect road edges especially because of unfavourable environmental conditions like rain, fog, night driving etc. [12]. The speed of vehicles and curvy roads can also result in off-road accidents, but here the study focuses on edge detection in unfavourable weather conditions. Different intelligent methods are used to avoid such problems by clearly identifying road edges.

398

H. Varun Chand et al.

Fig. 2 Architecture for parking space allocation system

Context: The estimation of ambient visibility range in the presence of fog is very difficult. A deep convolutional neural network approach has been used to enhance the awareness of driving weather conditions [13]. Objective: To provide driver awareness about the weather conditions and thereby reduce the safety risk. Method: A pre-trained deep convolutional neural network is used to extract the features of captured images. For visibility range estimation, d = − k1 ln(0.05) can be used. Support Vector Machines (SVM) algorithms are used to classify the visibility range. The AlexNet architecture is used as deep CNN, which has 5 convolutional layers, 3 softmax pooling layers and 3 fully connected layers. Results: The low visibility due to fog or bad weather can be detected using a deep CNNN method with an accuracy of 88.92. The system warns drivers and recommends the proper speed, which reduces the risk of accidents due to bad weather. Context: Advanced driver assistance systems help to assist drivers by providing safety and comfort. Computer vision also plays a crucial role in providing intelligence to these systems [14].

A Study on Parking Space Allocation and Road Edge Detection …

399

Objective: To identify the road edge in a foggy climate, using computer vision techniques. Method: Computer vision helps to reduce the cost and it provides more intelligence in detecting the road edges. If fogs are detected, canny edge detection algorithms are used which enhance the edges in the images. Also, the Hough line detector is used to estimate edge lines of the road images. The vanishing point of the image will be found after this step. Then road and sky segmentation and estimation is carried out. And finally, the visibility distance is estimated. Results: Visibility distance of roads are found using this computer vision technique. Context: For safe driving during fog, technological methods have to be applied to detect the edges of roads. Objective: A fog detection method is used to make driving more safe during foggy times. It selectively uses a de-fogging method if there is a presence of fog [15]. Method: The method is considered only for foggy images that are captured by the onboard camera of the vehicle. It uses de-fog and de-hazing techniques that prevent performance degradation due to decreased visibility. A fog detection algorithm is used to detect the presence of a foggy environment. The method is shown in Fig. 3. Because the image quality is strengthened and the fog from those images has been removed using the fog removal algorithm. The foggy images have low saturation and high value in the HSV color domain, which helps to identify the presence of fog. Results: The experimental result shows that this algorithm has a fog detection accuracy of more than 97% and it improves the image quality of existing de-fogging algorithms. Fig. 3 Overview of de-fogging scheme

400

H. Varun Chand et al.

Context: Identifying the edges of roads during all weather and illumination conditions are challenging. A driver video data mining technique is used to detect road edges [16]. Objective: Real-time driver video has been used for data mining using all weather and illumination conditions. Method: Data mining techniques have been applied to driver video data, to find out the qualitative and statistical distribution of road edges. Different weather and illumination conditions are also tested. It also uses the Bayesian method, a probabilistic approach to categories in on-road and off-road regions. Different illumination conditions like sunny and cloudy, rainy and wet roads, snowing and snow-covered roadsides, dark road surfaces, night and invisible roadside etc. can be detected with this approach. So, irrespective of the above-mentioned information, a compact road edge image can be generated from the video data. Results: The overall boundary accuracy for this method is 87%. By using the Bayesian method, prior knowledge helps to reduce the error of road edges.

4 Proposed Model The time wasted searching for parking slots and unclear visuals of road edges, vehicles, pedestrians etc. during unfavourable weather conditions etc. may result in traffic congestion. From the literature, a model has been proposed to provide solutions to such problems. The proposed model has two parts (1) for providing driver awareness about the ambient weather conditions and (2) for allocating parking lots. The proposed architecture for both are shown in Figs. 4 and 5 respectively. The proposed edge detection model identifies road edges in foggy weather conditions and provides assistance to the driver for accelerating or de-accelerating the vehicle’s speed. The on-board camera that captures real images has to be preprocessed using local binary patterns. These preprocessed images need to be fed into a deep convolutional neural network. Parallel to this, another method using canny edge detection has to be carried out to find the vanishing point of the image. Using both these methods will help to converge at a clear image, and thereby it can help to assist the driver in foggy and other unfavourable weather conditions. The proposed model for parking space allocation will collect images from onboard cameras. The model will identify current vacant slots around the roadside while the vehicle is moving and the details are updated on the cloud. This updated vacant slot information will be very useful for other drivers who might be searching for a slot around that area. The driver needs to request a parking slot along with the location where it is required. The request is processed in the cloud where updated information about the allocated and vacant slot is kept. Based on the request received, the cloud will provide a nearby vacant slot based on auction-based methods. The user or driver then opts for a V2G method, where the electric power of vehicles is stored

A Study on Parking Space Allocation and Road Edge Detection …

401

Fig. 4 Proposed model for road edge and fog detection

Fig. 5 Proposed model for parking space allocation

on to the charging unit near the parking slot. Nowadays, vehicles are moving towards pollution-free methods. So such future options are also added to this model.

402

H. Varun Chand et al.

5 Conclusion A model has been proposed to provide solutions to two major problems in VANET. With the advancement of new technologies, parking space allocation and road edge detection can be done with so much ease. Off-road accidents that occur due to poor judgment of road edges, especially during unfavourable environmental conditions, can be avoided with the help of technologies. The study focused on optimization of traffic by providing new cutting edge technology, to solve two major problems i.e., parking space identification and road edge detection. Road edge detections help to reduce off road accidents and thereby ensure smooth traffic on the road. The road edge detection indirectly influences traffic congestion. From this literature it’s clear that the new technologies focusing on parking space allocation and road edge detection provide a solution to traffic optimization during peak hours.

References 1. Afrin T, Yodo N (2020) A survey of road traffic congestion measures towards a sustainable and resilient transportation system. Sustainability 12(11):4660 2. Bosch E, Jhme U, Drewitz Jipp M, Oehl M (2020) Why drivers are frustrated: results from a diary study and focus groups. Eur Trans Res Rev 12:52 3. Parmar J, Das P, Dave SM (2020) Study on demand and characteristics of parking system in urban areas: a review. J Traffic Transp Eng (Engl Ed) 7(1):111–124 4. Ruan J, Liu B, Wei H, Qu Y, Zhu N, Zhou X (2016) How many and where to locate parking lots? A space–time accessibility-maximization modeling framework for special event traffic management. Urban Rail Transit 2:59–70 5. Ma Y, Liu Y, Zhang L, Cao Y, Guo S, Li H (2021) Research review on parking space detection method. Symmetry 13(128):1–18 6. Kuehn M, Hummel T, Bende J (2015) Analysis of car accidents caused by unintentional run off road. In: Proceedings of the 24th international technical conference on the enhanced safety of vehicles (ESV), Sweden 7. Hashimoto S, Kanamori R, Ito T (2013) Auction-based parking reservation system with electricity trading. In: 2013 IEee 15th conference on business informatics, Vienna, pp 33–40 8. Ji Z, Ganchey I, O’Droma M, Zhao L, Zhang X (2014) A cloud-based car parking middleware for IoT-based smart cities: design and implementation. Sensors 14(12):22372–22393 9. Safi QGK, Luo S, Pan L, Liu W, Hussain R, Bok SH (2018) SVPS: cloud-based smart vehicle parking system over ubiquitous VANETs. Comput Netw 138:18–30 10. Rane S, Dubey A, Parida T (2017) Design of IoT based intelligent parking system using image processing algorithms. In: 2017 international conference on computing methodologies and communication (ICCMC), Erode, pp 1049–1053 11. Chand V, Karthikeyan J (2020) Design and implementation of parking system using feature extraction and pattern recognition technique, In: Peter J, Fernandes S, Alavi A (eds) Intelligence in big data technologies—beyond the hype. advances in intelligent systems and computing, vol 1167, Singapore, pp 389–400 12. Jagerbrand AK, Sjobergh J (2016) Effects of weather conditions, light conditions, and road lighting on vehicle speed, vol 5. Springer Plus, p 505 13. Outay F, Taha B, Chaabani H, Kamoun F, Werghi N, Yasar A (2021) Estimating ambient visibility in the presence of fog: a deep convolutional neural network approach. Pers Ubiquit Comput 25:51–62

A Study on Parking Space Allocation and Road Edge Detection …

403

14. Bronte S, Bergasa LM, Alcantarilla PF (2009) Fog detection system based on computer vision techniques. In: 2009 12th international IEEE conference on intelligent transportation systems, St. Louis, USA, pp 1–6 15. Choi KY, Jeong KM, Song BC (2017) Fog detection for de-fogging of road driving images. In: 2017 IEEE 20th international conference on intelligent transportation systems (ITSC), Yokohama, Japan, pp 1–6 16. Wang Z, Cheng G, Zheng JY (2019) Road edge detection in all weather and illumination via driving video mining. IEEE Trans Intell Veh 4(2):232–243

Human Physical Activities Based Calorie Burn Calculator Using LSTM Jadhav Kalpesh, Jadhav Rushikesh, Kalbande Swaraj, Katta Rohan, and Rakhi Bharadwaj

Abstract Sensors can now recognize human physical activity with the recent technological advances. Accelerometers, gyroscopes, and magnetometers are some of the sensors embedded in smartphones. In today’s data-driven world, human activity recognition is important in a variety of fields, including medical applications, fitness tracking, human survey systems, and so on. This research study analyzes the data obtained from mobile sensors such as gyroscopes, accelerometers, linear accelerometers, and magnetometers. Further, the proposed model predicts the human activity by using the data collected from a mobile sensor. Sitting, standing, jogging, and other such activities can also be tracked. Calorie-Meter is an Android application that calculates the calories burned while engaging in such activities. Using the application’s predicted activity, the user’s calories burned and calorie deficiency can be calculated. This research study proposes the utilization of Long Short-Term Memory (LSTM) and a Neural Network (NN) technique for predicting the human activity based on sensor data. Keywords Human activity recognition · Sensors · Calorie-meter · Calorie deficit · Long short-term memory

J. Kalpesh · J. Rushikesh · K. Swaraj · K. Rohan · R. Bharadwaj (B) Department of Computer Engineering, Vishwakarma Institute of Technology, Pune, India e-mail: [email protected] J. Kalpesh e-mail: [email protected] J. Rushikesh e-mail: [email protected] K. Swaraj e-mail: [email protected] K. Rohan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_31

405

406

J. Kalpesh et al.

1 Introduction Human Activity Recognition (HAR) systems are becoming increasingly popular in the fields of fitness and medical applications. HAR is used in medical applications to care the elderly population by tracking their health records. A magnetometer is used to define the pole position in this project; an accelerometer is used to measure linear acceleration; a gyroscope is used to detect deviation of an object from its desired orientation. Henceforth, the human activities can be predicted by using these sensors. In this project, the dataset used for the training and testing a model is created by using the mobile application named “Sensor Logger” and different actions like running, standing, and climbing stairs were recorded for creating an input dataset. After generating the data, it gets divided into training and testing datasets and the model is then trained by using the Long Short-Term Memory (LSTM) algorithm. This model can now accurately predict the user’s current activity. A specific amount of calorie burning rate (Cal/hr) is assigned to every specific activity. Using this, the number of calories burned by that activity in a given amount of time is calculated by providing an exact number of calorie deficit count (Calorie burnt—Calorie intake). The above application allows Android users to easily plan the weight-loss diet plans and exercises.

2 Related Work Researchers [1] have recently considered the approach of data mining for human activity recognition. Here, big data has been used and further a dataset is created by using 30 test subjects. For the dataset readings, the test subjects used waist-mounted mobile phones for recording and analyzing the user’s day-to-day activities. The dataset considered here includes a variety of age groups ranging from 19 to 48 based on many attributes and instances. The dataset contains various physical activities, which are being recognized. This research study has attained an accuracy of 89%. Both the approaches of human activity recognition i.e., using sensors and human activity recognition using images are considered [2]. Demrozi et al. [2] have carried out a brief existing literature survey by considering 46 papers from various recent conferences. The papers include more information about the different CML methods of ML. It has enlisted a brief information about sensor fusion techniques. The sensor fusion techniques are used to overcome the errors of environmental noise. In normal HAR, the accelerometer can measure the speed accurately but it can be deceived by different angular movements. On the other hand, this research study includes a gyroscope to measure the angular movements, which can be deceived by some linear movements and further to avoid this problem, sensor fusion is used. In [3], authors have used all 6 basic human activity recognition movements in the dataset. The 6 different signals obtained are then stored as time series and recorded

Human Physical Activities Based Calorie Burn Calculator Using LSTM

407

with a sampling rate of 50 Hz. Dataset used here is generated from a gyroscope and accelerometer. The authors in [4] have proposed a novel method by using physical human activity and energy expenditure estimation in humans with the smartphone application. Also, the author has included an optional heart rate monitor. This method detects the movement of a device by using sensors, which are integrated into smartphones. This method will recognize the movements like linear acceleration and direction of magnetic field by using magnetometer etc. In [5], the researchers have presented an application, where the innovative method is used for performing activity recognition and analyzing the expenditure of human energy. The proposed application uses data from Smartphone and additionally an optimal heart rate monitor is used to predict the energy expenditure. In [6], a novel method has been developed for performing human activity classification by using a Support Vector Machine (SVM). They used accelerometers and gyroscope sensors for data collection. Usage of SVM ensemble techniques can leverage an accuracy of 99.1%, sensitivity of 99.6% and specificity of 98.7%. In [7], a new sensor device has been proposed by using accelerometer vibrations. User can wear sensor device to monitor the physical activities. This paper explains the prevention or detection of human activity by using artificial intelligence, conduction, assessment of rehabilitation exercises, and monitoring of neurological disease progression. The author of [8], gave a brief idea of support-vector networks. With the explanation of two-group classification problems and the idea of the machine, which implements conceptually. Also, they constructed the feature space in a linear decision surface. Some special properties of decision surface ensure the high generalization ability of the machine learning model. In [9] the author has conducted a survey on smartphone apps which are used for calorie counting and compared its effect on nutritional awareness and lifestyle modifications. The proposed survey has compared the top 20 apps available on google play store in android applications. The quality of these applications was analyzed using a 55-point scoring system. The characteristics such as accuracy of content, the usability of the interface, standards used, and database source were compared and analyzed in this survey. In [10], the paper explains the work on suspicious human activity detection using two models, which are trained with a supervised CNN and unsupervised CNN, which doesn’t need any records. It uses 250–300 frames of videos in mp4 format as a dataset for performing different activities and successfully predicts the suspicious outdoor human activities which are tracked by an ordinary CCTV camera. In [11], the authors proposed a novel method to recognize human activity using object information for the detection of different activities. They used activity theory to propose a new technique for activity recognition. For activity theory, authors needed calculation and statistical models i.e., Hidden Markov Model is used. In this paper physiological sensors, cameras, and RFID sensors are used for inputs. Penalized Naive Bayes Classifier is used for different approaches.

408

J. Kalpesh et al.

In [12], the authors analyzed the difference between the conversation of humans and chatbots. There is various open chat networks where bots are found abusing human beings. This paper studies these bots. Bots are spreading spam and malware through these networks. The chats are distinguished based on entropy i.e., the amount of time between two text messages. In the [13] paper, the authors use sensors on the body to provide data input. The daily human activities are classified into four categories, they are stationary i.e., not moving (in resting position), light ambulatory i.e., slow walking, intense ambulatory i.e., fast walking or jogging, and abnormal activities like sudden falling of subjects due to a sudden change in environment. The support vector machine based learning and decision tree techniques are used. The accuracy score found is 82.76%, 69.56%, 70.56%, and 60.15% for activities such as Stationary, light ambulatory, intense ambulatory, and abnormal respectively. In [14], the authors have analyzed different classification techniques like CNN, random forest, and support vector machine. Datasets used here are publicly available. This dataset includes 6 regular human activities calculated with the help of smartphone accelerometers. Activities include walking, jogging, running, sitting, standing, climbing upstairs, and climbing downstairs. The accuracy of different algorithms are calculated, and it was proven that CNN produces the highest accuracy, better than SVM and Random Forest. In [15] the author used various benchmark datasets to recognize human activity. They selected machine learning and deep learning-based solutions. Also explain the data collection, data pre-processing, and database structure. Importantly they explain that for sensory signals of imaging, a convolutional neural network has been used and for raw sensory, Convolution Neural Network (CNN), Long Short-Term Memory (LSTM), and hybrid models are used. In [16], authors Harris and Benedict have briefly stated their equation for calculating Basal Metabolic Rate (BMR) after a series of experiments and observations. Normal human individuals of both sex and different ages were used as subjects in 1919. In [17], Mifflin has proposed a new equation to predict Basal Metabolic Rate (BMR) in healthy individuals. Harris and Benedict’s equation for BMR was made more accurate after studies on 498 healthy human subjects. It has then corrected to perform the overestimation of REE by Harris and Benedict.

3 Methodology 3.1 Process Flowchart Data collection is the first step to start any machine learning model. Data was collected by using the ‘Sensors Logger App’. Using this app, the data has been collected from mobile sensors for performing various activities and in various orientations. Data

Human Physical Activities Based Calorie Burn Calculator Using LSTM

409

Fig. 1 Process flowchart

collected from the app is stored in CSV format. Further, it is divided into 80% as training dataset and 20% as testing dataset. Training data has trained the proposed model by using the Long Short-Term Memory (LSTM) algorithm and a model was generated. The pb Model generated here was then exported into an Android application, where it is used to predict activity. Using this we can calculate calories burnt, the calorie deficiency of the user, and show physical activity distribution on the pie chart at a particular time, which will be discussed in Fig. 1.

3.2 Data Collection Data collection is the first phase of any machine learning project. In this project, the data was collected using the help of the sensors used in the smartphone itself. The Smartphone used was common in all trials. The smartphone model was Xiaomi’s Redmi Note 7 Pro. Ten different volunteers participated in this experiment. The smartphone was attached in five different orientations as follows: 1. 2. 3. 4.

left pocket right pocket wrist arm

410

J. Kalpesh et al.

5. belt. During each orientation, all seven different activities were recorded for seven minutes. The activities were as follows: 1. 2. 3. 4. 5. 6. 7.

standing sitting running walking jogging stair climbing upstairs stair climbing downstairs.

Sensors of smartphones used to collect data: 1. Accelerometer: An accelerometer uses gravity as a reference to detect the orientation of the phone as well as it can detect motion. When we view the phone in portrait mode or landscape mode, it is detected by the accelerometer, as it detects orientation change. The accelerometer is an electromechanical device that measures acceleration caused due to various natural forces like gravitational force, magnetic force, etc. Accelerometer uses piezoelectric crystals like quartz whose small crystal is attached to a small mass. When the accelerometer moves, that mass squeezes quartz crystal which generates a small voltage that is used to calculate acceleration. 2. Linear accelerometer: This sensor is used in smartphones for getting the acceleration of the phone without considering gravity. This provides data along all three axes in the x, y, and z. 3. Gyroscope: To measure the device’s rate of rotation in radian per second we use a Gyroscope. It measures them in all 3 axes of the device. Gyroscopes used in smartphones are Vibration Gyro Sensors that measure angular velocity using Epson’s double-T structure crystal element which is a vibrating object. Coriolis force applied on this vibrating object help measure angular velocity in radian per second. 4. Magnetometer: Smartphones come equipped with a magnetometer so that your phone can sense its orientation in space to determine your location for Magnetic North. All four sensors were used for data gathering. An application from Google Play Store ‘Sensors Logger’ was used for data collection. This application records the data gathered by the different sensors and converts it into CSV format. 50 Hz is the frequency of data points that are collected using this application. A total of 630,000 samples were collected. 63,000 samples per person. For each activity, 9000 samples were recorded in 5 different orientations per person.

Human Physical Activities Based Calorie Burn Calculator Using LSTM

411

3.3 Data Cleaning and Data Labeling Removing unusable data from all the datasets was an important factor before labeling the data i.e., removing the starting and ending rows of columns in the CSV files due to unnecessary observations. The observations that were recorded during the transitions from one activity to another were also removed. 1800 sample points per activity per orientation per person were recorded at last. The data labeling was done manually in the CSV files themselves. This data was labeled into the seven activities mentioned above. The final part was to merge all the different CSV files of a single volunteer into one single CSV file for easier access and model generation. Figures 2, 3, 4 and 5 were generated by using MS Excel from the final Dataset. Fig. 2 Accelerometer sensor pattern of person 1. x-axis: a point in time, y-axis: m/s2

Fig. 3 Linear Accelerometer sensor pattern of person 1. x-axis: a point in time, y-axis: m/s2

Fig. 4 Gyroscope sensor pattern of person 1. x-axis: a point in time, y-axis: rad/s

412

J. Kalpesh et al.

Fig. 5 Magnetometer sensor pattern of person 1. x-axis: a point in time, y-axis: µT (micro-Tesla)

3.4 Patterns in sensor’s Data Recorded in the Activity Many algorithms are used for Time Series Forecasting, they are ARIMA, Prophet, and LSTM. While dealing with a large dataset where lakhs of samples are needed to be trained, LSTM suits best among all. Also, LSTM is easy to use and implement using Python language. It is known that LSTM works on patterns and recognizes them. The challenge of a basic feed-forward neural network is that it can’t remember patterns, in short, it has no memory. A neural network that remembers patterns over time series given durational time lags is called Long Short-Term Memory (LSTM). Data of sensors which was recorded had some patterns which we can see in graphs. Each data point in the dataset recorded, corresponds to a point in time. From the above graphs generated by using MS Excel, we can observe that every reading done by four sensors has a pattern that is a function of time intervals. LSTM’s work is to identify this pattern and predict the result of human activity. LSTM is not sensitive to the length of the time gap, which is not the case in RNNs, sequence learning models, and other Markov models. This is one of the significant advantages of using LSTM. For example, if we want to predict something after 10,000 intervals of time, RNNs, and Markov models will forget but LSTM will remember it. But in the case of short intervals like 100, RNNs may act as normal LSTMs.

3.5 Proposed Methodology for Activity Recognition Using LSTM In theory, RNNs are capable of predicting using long-term dependencies, but in practice, it doesn’t seem to be able to learn them. Hochreiter and Bengio discovered why RNNs cannot learn long-term dependencies in practice. In 1997 Hochreiter and Schmidhuber introduced LSTM and it was refined and popularized later [18]. LSTM is a type of recurrent neural network. We can also define it as a Recurrent Neural Network that can remember things over the gap of a certain length or intervals. LSTM is used when we must predict something which is the function of time series or in other words time series forecasting (Fig. 6).

Human Physical Activities Based Calorie Burn Calculator Using LSTM

413

Fig. 6 RNN structure

Fig. 7 LSTM cell

RNN cell has two inputs, which help in prediction, that is output of the last state and input given to cell at time t. No information of the past can be retained in a hidden state in the case of RNNs (Fig. 7). To remember long-term memory, we need another state in the RNN cell, self-state i.e., long term memory besides hidden state i.e. short-term memory. LSTM cell as we observe is similar to RNN cell and has a recurring structure. In RNN, there is only one Neural Network layer of tanh. But this is not true in the case of LSTM a special way of layers are present, where it can have up to four neural networks in simultaneous interaction (Figs. 8 and 9). The sigmoid activation layer and tanh activation layer are used. Sigmoid function varies between 0 and 1 while tanh function varies between −1 and +1. In Sigmoid function is an S-shaped curve that takes real-valued input. tanh function is mostly used in RNN for NLP and Speech Recognition. It varies from −1 and +1 which helps in the backpropagation process (Figs. 10 and 11). Gates have a sigmoid function layer. Gate also has a pointwise multiplication operation. Three gates are present in Long Short-Term Memory. LSTM remembers using a long-term state. If that long-term state needs to be changed, it is done using Forget Gate. The role of Forget Gate is when it encounters certain values during prediction, it must know, it must discard previous long-term memory, and store the new word in this state. To store this new state, we have an input gate. Using Sigmoid and tanh function on Hidden state and new word, cell replaces a new word in long

414

J. Kalpesh et al.

Fig. 8 tanh function activation layer

Fig. 9 Sigmoid function activation layer

term state which means we are adding the memory of new value cell read. The third one is the Output gate, where we do the weighted sum of hidden state and value read and then use the sigmoid function on it. Later we multiply it with tanh of new long-term memory to get a new hidden state (Figs. 12, 13 and 14). Step-by-step LSTM journey. Step 1: Forget Gate will decide which information that entered cell is important and discard which is not needed. It takes ht −1 and text as input and outputs number ∈ [0,1]. See Fig. 12.

Human Physical Activities Based Calorie Burn Calculator Using LSTM

415

Fig. 10 Short term memory cell in traditional RNN

Fig. 11 Short-term and long-term memory

Fig. 12 Forget gate

f t = σ (W f .[h t−1 , xt ] + b f )

(1)

here ft is an output of forget gate, W f is the weight of forget gate neuron cell, ht −1 is previously hidden state input and x t is current input, bf is biased. The sigmoid

416

J. Kalpesh et al.

Fig. 13 Input gate

Fig. 14 Output gate

function gives a value between [0, 1] and if its 0 means we must forget the previous memory. Step 2: Information that needed to be stored in the cell state is decided by LSTM, which has 2 parts. The first one is the Sigmoid function which is also called as Input Gate Layer. The job of the Input Gate layer is to decide which value to update. The second is the tanh layer. It calculates a new candidate value’s vector Ct which could be added to the state. i t = σ (Wi .[h t−1 , xt ] + bi )

(2)

here it is an output of the sigmoid input gate, W i is the weight of the sigmoid input gate neuron cell, ht −1 is previously hidden state input and x it is the current input and bi is bias for input gate. Ct = tanh(Wc .[h t−1 , xt ] + bc )

(3)

Human Physical Activities Based Calorie Burn Calculator Using LSTM

417

here Ct is a temporary new state which is output of tanh input gate, W c is the weight of tanh input gate neuron cell, ht −1 is the previously hidden state input it is the current input and bcthe is bias for the input gate. Now we update the old cell state C t −1 . We multiply old state with f(t), forgetting things we decided to forget earlier, then we add i(t) ∗ Ct . Ct = f t ∗ Ct−1 + i t ∗ Ct

(4)

here C t is new Cell State, f t is output of forget gate, C t −1 is the previous Cell State, it is the output of the sigmoid input gate and Ct is the temporary new state which is the output of the tanh input gate. Step 3: Finally, LSTM needs to decide what the output is going to be. It will be a filtered version and based on cell state. We use sigmoid layer that determines the part of the cell we will return as output. Later apply tanh on cell state to multiply it with output of sigmoid gate earlier, so we output the parts we decided to. ot = σ (Wo .[h t−1 , xt ] + bo )

(5)

here it is output of sigmoid gate in Output Gate, W o is weight of output gate neuron cell, ht −1 is previous hidden state input and x t is current input, bo is bias. h t = ot ∗ tanh(Ct )

(6)

here ht is new hidden state which is output of Output Gate, ot is an output of Output Gate and C t is the new Cell State which is output of Input Gate. Using LSTMs, we know patterns in datasets captured by sensors, using which we can do time series forecasting.

3.6 Training of Model and Testing of Model For building the LSTM model, we considered 7 classes and 32 hidden units, and for activation function we used the ReLU function and kernel initializer as orthogonal. The model is trained for 30 epochs and the batch size is 1024. The training and testing dataset is divided into 80:20 ratio, 504,000 samples are used for training of model while 126,000 are used for testing the model. The time needed for training and evaluation takes 38 s and the average time taken for each epoch is around 1 s 33 ms. The accuracy of the model is 98.17. For checking the performance of model, the confusion matrix is created which is shown in Fig. 15.

418

J. Kalpesh et al.

Fig. 15 Confusion matrix

3.7 Calculating Calorie Deficit Using the Predicted Graph Using activity prediction done by LSTM model, it can now approximate the calorie burndown a person is having. Calories are units used to measure energy expenditure. The amount of calories energy our food or drink is measured in calories. A common unit used while dealing with energy dynamics of the human body is kcal. But as kcal is so commonly used in this field, it’s commonly called Calorie. The “calorie” we refer to in food is kilocalories. One kilocalorie = (uppercase C) one Calorie. In scientific terms, a kilocalorie is basically amount of heat we require to raise the temperature of 1 kg of water by 1 °C. Every person, depending on his age, weight, height, and gender has some default energy expenditure in a day, which happens even if you do no physical activity. This is called Basal Metabolic Rate. The basic metabolic rate is equal to minimum number of calories a body need in non-working state. This can be estimated using equations. One of these equations is the Harris-Benedict equation [16, 17], which is a formula that takes one’s weight, age, height, gender as input to find BMR or Total Energy Expenditure. BMR calculators are available online. Men: B M Rm = 88.362 + (13.397 × W ) + (4.799 × H )−(5.677 × A)

(7)

where W is weight in kg, H is height in cm, A is age in years. Women: BMRw = 447.593 + (9.247 × W) + (3.098 × H)−(4.330 × A)

(8)

Human Physical Activities Based Calorie Burn Calculator Using LSTM

419

where W is weight in kg, H is height in cm, A is age in years. Calorie burning rate (Cal per hour) for person with weight of 130 lb in activities we are recognizing using model, as per [10] are: 1. 2. 3. 4. 5. 6. 7.

standing: 165 Cal per h sitting: 148 Cal per h running: 472 Cal per h walking: 224 Cal per h jogging: 372 Cal per h stair climbing upstairs: 472 Cal per h stair climbing downstairs: 177 Cal per h.

Model will predict activities done by person and then using Calorie burning rate will calculate calories burned. To calculate Calorie deficit, one must consider calories taken as input. It includes daily items we eat and drink. There are various mobile apps which keep track of these calories we intake.  (ti ∗ Cal Ai ) (9) Calorie bur nt = B M Ru + here t i is time for ith activity, CalAi is Calorie burning rate in Cal per hour for ith activity, BMRu is Basal Metabolic Rate of a particular user. Calorie de f icit = Calorie bur nt − Calorie intake

(10)

Calorie burnt is energy expenditure due to various physical activities and Calorie intake in energy a person takes after consumption of food. Units are Calories for every term in equation. Users must input Calories intake into the application developed so that we can calculate Calorie deficit. Using the above technique, we can run an application for a particular amount of time or a day and calculate Calorie deficit for that time or complete day.

4 Result and Conclusion This paper is proposing new Calorie-Meter which is name of android application using Smartphone sensors that are used to gather data and long short-term memory is used for model generation. There are 7 activities that are classified using this model are: standing, running, walking, upstairs, downstairs and jogging. ‘Sensors Logger” was application used to collect data from smartphone sensor. The signals were received in form of x, y and z coordination system in the 3-dimensional space form every sensor. Data is logged from smartphone sensors and signals given by them while performing the above 7 activities.

420

J. Kalpesh et al.

This application takes input from the users such as age, height, and weight. model predicts the user’s activities by the person and calorie burnt is calculated. The application tells the user how much calories they have burnt in a day and showcases various activities in pie chart representation.

4.1 Accuracy of Model Accuracy of model built is 98.13% and loss is 0.0707.

4.2 Calorie-Meter on Andorid App Key part of this project is the Android application implementation which carries out live activity detection of a person carrying a smartphone. Design is such that it’s easy to understand and user-friendly. Basic user input like weight, height, age, and gender are taken and stored locally in the application. The Calorie-Meter app basically provides users a report of calories they burned and physical activity levels using a Pie chart (screenshots in Figs. 16 and 17). Using information taken from users, app estimates calories burnt in during activities carried out. User have to start timer of the app. From that point onwards every activity of the user will be recorded along with the time of each activity performed. Once the user stops the timer, a Pie chart is displayed (as shown in Fig. 17) which shows activity distribution in the ratio of time it is performed. The proposed study has successfully recorded the development of a smartphone Android application that performs live detection of user’s physical activities. This app is different from previous works on activity recognition in the following: (1) User interaction needed after setup is close to zero; (2) It relies solely on sensors of smartphones which are available with low-end smartphones too and don’t require additional sensing hardware; (3) It recognizes seven different human activities, including walking, running, climbing stairs, descending stairs, sitting, standing, jogging; (4) App is user friendly and easy to understand, which include Graphical representation of data gathered; (5) Accuracy of Calorie-Meter app built is 98.13% which is reliable; (6) App can also calculate Calorie deficit. This functionality is not present in many apps in the marketplace. Application lets user monitor their daily physical activity and enable them to make healthier choices and make good and healthier habits. The application is targeted to encourage people toward a healthy lifestyle.

Human Physical Activities Based Calorie Burn Calculator Using LSTM

421

Fig. 16 Human activity recognition dashboard and start-stop timer

5 Future Directions This project has tremendous scope for future research, wherein the larger dataset can result in increasing the accuracy of the model proposed. Moreover, this will help to maintain the user health after analyzing their daily activities and maintaining calorie goals. The proposed application can be further extended to (1) recognize more physical activities, (2) extra features can be added to provide users with the information about their heart rate, obesity levels, and carbon footprints, and (3)

422 Fig. 17 Pie chart of human activities in particular time, Calorie deficit calculated, and Calorie burnt

J. Kalpesh et al.

Human Physical Activities Based Calorie Burn Calculator Using LSTM

423

develop additional functionality, which uses stored historical information of user’s daily activities and help user to make appropriate lifestyle choices, which are healthier and smart. Acknowledgements We would especially like to thank Vishwakarma Institute of Technology (VIT), Pune for providing us with a platform for producing such a wonderful project. We would like to acknowledge our honorable Director Prof. Dr. Rajesh M. Jalnekar for his inspiration and guidance. Also, the Head of the Department of Computer Engineering Prof. (Dr.) Sandeep Shinde for his valuable support, help, and time. Valuable support and inspiration were given to us by our project guide Prof. Rakhi Bharadwaj for writing this paper.

References 1. Ann OC, Theng LB (2014) Human activity recognition: a review. In: 2014 IEEE international conference on control system, computing and engineering (ICCSCE 2014), pp 389–393. https:// doi.org/10.1109/ICCSCE.2014.7072750 2. Demrozi F, Pravadelli G, Bihorac A, Rashidi P (2020) Human activity recognition using inertial, physiological and environmental sensors: a comprehensive survey. IEEE Access 8:210816– 210836. https://doi.org/10.1109/ACCESS.2020.3037715 3. Bulbul E, Cetin A, Dogru IA (2018) Human activity recognition using smartphones. In: 2018 2nd international symposium on multidisciplinary studies and innovative technologies (ISMSIT), pp 1–6. https://doi.org/10.1109/ISMSIT.2018.8567275 4. Cvetkovi´c B, Janko V, Luštrek M (2015) Activity recognition and human energy expenditure estimation with a smartphone 5. Anjum A, Ilyas MU (2013) Activity recognition using smartphone sensors. In: 2013 IEEE 10th consumer communications and networking conference (CCNC), pp 914–919. https://doi.org/ 10.1109/CCNC.2013.6488584 6. Hardiyanti N, Lawi A, Diaraya, Aziz F (2018) Classification of human activity based on sensor accelerometer and gyroscope using ensemble SVM method. In: 2018 2nd East Indonesia conference on computer and information technology (EIConCIT), pp 304–307. https://doi.org/ 10.1109/EIConCIT.2018.8878627 7. Caya MVC, Yumang AN, Arai JV, Niñofranco JDA, Yap KAS (2019) Human activity recognition based on accelerometer vibrations using artificial neural network. In: 2019 IEEE 11th international conference on humanoid, nanotechnology, information technology, communication and control, environment, and management (HNICEM), pp 1–5. https://doi.org/10.1109/ HNICEM48295.2019.9072850 8. Cortes C, Vapnik V (1995) Support-vector networks. Mach Learn 20(3):273–297 9. Banerjee P, Mendu VVR, Korrapati D, Gavaravarapu SM (2020) Calorie counting smart phone apps: effectiveness in nutritional awareness, lifestyle modification and weight management among young Indian adults. Health Inf J 10. Bhambri P, Bagga S, Priya D, Singh H, Dhiman HK (2020) Suspicious human activity detection system. J IoT Soc Mob Anal Cloud 2(4):216–221 11. Yang J, Lee J, Choi J (2011) Activity recognition based on RFID object usage for smart mobile devices. J Comput Sci Technol 26:239–246 12. Smys S, Wang H (2021) Naïve Bayes and entropy based analysis and classification of humans and chat bots. J ISMAC 3:40–49. https://doi.org/10.36548/jismac.2021.1.004 13. Ali SE, Khan AN, Zia S, Mukhtar M (2020) Human activity recognition system using smart phone based accelerometer and machine learning. In: 2020 IEEE international conference on industry 4.0, artificial intelligence, and communications technology (IAICT), pp 69–74. https:// doi.org/10.1109/IAICT50021.2020.9172037

424

J. Kalpesh et al.

14. Khatun MA, Yousuf MA (2020) Human activity recognition using smartphone sensor based on selective classifiers. In: 2020 2nd international conference on sustainable technologies for industry 4.0 (STI), pp 1–6. https://doi.org/10.1109/STI50764.2020.9350486 15. Ramanujam E, Perumal T, Padmavathi S (2021) Human activity recognition with smartphone and wearable sensors using deep learning techniques: a review. IEEE Sens J 21(12):13029– 13040. https://doi.org/10.1109/JSEN.2021.3069927 16. Harris JA, Benedict FG (1918) A biometric study of human basal metabolism. Proc Natl Acad Sci U S A 4(12):370–373. https://doi.org/10.1073/pnas.4.12.370. PMID: 16576330; PMCID: PMC1091498 17. Mifflin MD, St Jeor ST, Hill LA, Scott BJ, Daugherty SA, Koh YO (1990) A new predictive equation for resting energy expenditure in healthy individuals. Am J Clin Nutr 51(2):241–247. https://doi.org/10.1093/ajcn/51.2.241. PMID: 2305711 18. Hochreiter S, Schmidhuber J (1997) Long short-term memory. Neural Comput 9(8):1735–1780. https://doi.org/10.1162/neco.1997.9.8.1

Alternate Tiny Encryption Algorithm: A Modified Tiny Encryption Algorithm for Improved Data Security Mehak Gupta , Nimit Agrawal , and Manas Ranjan Prusty

Abstract In this era of Industry 4.0, securing file data is very crucial in today’s environment with respect to data transfer of Internet of Things (IoT) devices. Over the years with the evolution of technology and file storage systems, many algorithms have been used for encryption and decryption processes for securing the file data, each with its methodologies, advantages, and limitations. An efficient algorithm has very few limitations thus making it a top choice for usage. In this paper, we have proposed a symmetric key cryptographic algorithm called the Alternate Tiny Encryption Algorithm (ATEA) focusing on a strong approach for safekeeping of the file data and minimizing the weak points of the existing Tiny Encryption Algorithm (TEA) and Extended Tiny Encryption Algorithm (XTEA). The Alternate Tiny Encryption Algorithm (ATEA) is a Feistel cipher that utilizes mixed algebraic group operations. The algorithm is simple enough to incorporate into practically any computer program and can be quickly translated into a variety of languages. It uses a unique key generation technique making the encryption of file data more secure. Keywords Cryptography · Tiny encryption algorithm · Encryption · Decryption

M. Gupta (B) · N. Agrawal · M. R. Prusty School of Computer Science and Engineering, Vellore Institute of Technology, Chennai, Tamil Nadu 600127, India e-mail: [email protected] M. R. Prusty Centre for Cyber Physical Systems, Vellore Institute of Technology, Chennai, Tamil Nadu 600127, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_32

425

426

M. Gupta et al.

1 Introduction With the increase in digitalization of data, though managing and assembling data has become quite a lot easier, it has however increased the risks of security breach. Some vital information transferred within a workspace, across businesses, between divisions of an organization and other external bodies and institutions occasionally falls into the hands of illegitimate parties who could alter the information. If no security measures are adopted, such data and other confidential information will likely be subject to threats such as manipulation, theft, corruption, dispute, and break-in, resulting in substantial harm to the individual or organization. Alongside digitalization, the inception of Industry 4.0 has made a significant role in automation which involves various IoT devices. Data transfer in such devices needs to be upgraded with better and faster secure policies. Hence, this motivates researchers in improving the encryption and decryption algorithms and processes too. The idea of encryption/decryption algorithms is used where encryption is scrambling the plain text into cipher text and vice versa for decryption. For this process, we can use any of the two types of cryptographic keys present: symmetric or asymmetric. Symmetric or secret-key encryption algorithms encrypt and decrypt using the same symmetric key. Whereas asymmetric or public-key encryption algorithms encrypt and decrypt data two different but related keys. Symmetric algorithms are substantially faster than asymmetric algorithms and can accommodate thousands of keys with very little computing overhead [1]. Therefore, for our algorithm, encoding/decoding is done with the help of a secret key. A block cipher’s strength can be evaluated using characteristics like avalanche, completeness, and statistical independence [2]. The Butterfly Effect, commonly referred to as the Avalanche Effect, is a significant characteristic of cryptographic algorithms. It implies that altering just one bit in the plain text or the key should cause a significant change in the result. We verified the effectiveness of our proposed algorithm by calculating the avalanche effect. Completeness is a required feature of encryption, which means that each bit of the cipher text/output block must depend on each bit in the plaintext [3]. ATEA (Alternate Tiny Encryption Algorithm) is designed as a Feistel cipher. A Feistel network employs a round function that takes two inputs, a data block and a key, and gives one data block-sized output. The round function is applied to the half of the data to be encrypted in each round, and the output is exclusive ORed with the other half. This process is repeated a specific number of times, with the encrypted data as the final outcome. The process of decryption in the Feistel cipher is almost similar but not exactly the same. Instead of starting with the plaintext, the cipher text is sent into the Feistel structure and then the process thereafter is the same as that of encryption. The only difference in the case of decryption is that the sub keys used in encryption are utilized in the reverse order [4].

Alternate Tiny Encryption Algorithm: A Modified Tiny Encryption …

427

2 Related Work There are many algorithms developed for cryptography methods. Mentioned below are the previously existing methods which have been taken into account for developing our algorithm [5]. David Wheeler and Roger Needham developed the Tiny Encryption Algorithm (TEA). It is a 64-round Feistel block cipher with 64-bit block size and 128-bit key that consist of four 32-bit words K[0], K[1], K[2] and K[3]. It uses XOR (Exclusive OR), ADD, SUB and SHIFT operations. Although TEA is simple and cryptographically strong, it still has some weaknesses. Each key is equivalent to three others resulting in an effective key size of only 126 bits [6] which led to an attack on the gaming console of Microsoft’s Xbox where TEA was used as a hash function [7]. It lacks a key schedule algorithm, making it vulnerable to a related-key attack which necessitates 223 chosen plaintexts under a related-key pair and has a 223 time complexity [8]. Because of these flaws, the XTEA (Extended Tiny Encryption Algorithm) cipher was designed. The first version of XTEA was presented in 1997 by Wheeler and Needham, three years after TEA was first introduced [9]. XTEA is a block cipher that employs a 64 round Feistel structure with a 128-bit key as in TEA. Changes to the key schedule algorithm were introduced in XTEA. Instead of having the subkeys placed in a predetermined order, they were introduced as subkey A and subkey B, which are chosen using two bits of the variable “sum”. In addition, a shift of 11 was also added to the key schedule to give the subkeys a more erratic pattern. Rearranged addition, shift and XOR operations are among the other modifications that were made in XTEA.

3 Proposed System Encryption/Decryption algorithms deal with the data set to be encrypted, a set of keys to encrypt the data set into ciphertext and further, the keys are to be used to decrypt the ciphertext into the original data set. The proposed algorithm—ATEA is a cryptographically strong algorithm to protect file data. It is an improved version of TEA [5]. ATEA is a Feistel structure having 32 rounds (16 cycles) for each encryption and decryption. The proposed algorithm consists mainly of four modules—Key generation, Subkey generation, Encryption and Decryption.

428

M. Gupta et al.

3.1 Key Generation The encryption key will be produced automatically using the user’s random mouse pointer coordinates on the screen, making each key unique. Then the key is further converted into a hash key of 128 bits using a cryptographic hash function called salted MD5 (Message-Digest algorithm 5) [10]. MD5 is a simple and fast hashing algorithm that is added with salt to force its uniqueness, improve complexity without increasing user requirements, and protect against password attacks [11]. Salting a hash means adding a random string of letters and numbers, called a salt, before hashing it [12]. Hashing is a one-way process that forms a fixed-length binary sequence that is difficult to invert computationally. Hence, the attacker cannot retrieve the master key from the hash key, which makes the system more secure.

3.2 Subkey Generation The hash key generated using salted MD5 can now be used to generate subkeys that are to be inputted for encryption and decryption processes. We have proposed a new technique for this purpose that is easy to understand and implement. The proposed technique generates 64 unique subkeys which ensure that the subkeys are uniquely chosen for each cycle. First, the 128-bit hash key is divided and split into four 32-bit pieces- K[0], K[1], K[2] and K[3]. Then an array P of size 16 (P0, P1, …, P14, P15) is initialized with elements having 32-bit value each. Now the subkeys (S[0], S[1], …, S[62], S[63]) are generated by XOR operation between the elements of the P array and the four hash key pieces, as shown in Fig. 1, which will be used for the encryption and decryption process.

3.3 Encryption The file text to be encrypted is first split into two halves vv[0] and vv[1]. ATEA contains two rounds for each encryption or decryption cycle. In total, encryption contains 32 rounds that are 16 cycles. Round one and following odd rounds operate on vv[0]. Round two and following even rounds operate on vv[1]. The round function which consists of some basic operations like Bitwise shift, XOR and addition is applied to one half using subkeys. The round function output is added to the other half creating repeated mixing of the data and all the bits of the subkey. The two halves are then exchanged. The same pattern is observed in each round using unique subkeys generated in the previous step. Figure 2 shows an ATEA encryption cycle consisting of two rounds.

Alternate Tiny Encryption Algorithm: A Modified Tiny Encryption …

429

S[0] = P0

K[0]

S[16] = P4

K[0]

S[32] = P8

K[0]

S[48] = P12

K[0]

S[1] = P0

K[1]

S[17] = P4

K[1]

S[33] = P8

K[1]

S[49] = P12

K[1]

S[2] = P0

K[2]

S[18] = P4

K[2]

S[34] = P8

K[2]

S[50] = P12

K[2]

S[3] = P0

K[3]

S[19] = P4

K[3]

S[35] = P8

K[3]

S[51] = P12

K[3]

S[4] = P1

K[0]

S[20] = P5

K[0]

S[36] = P9

K[0]

S[52] = P13

K[0]

S[5] = P1

K[1]

S[21] = P5

K[1]

S[37] = P9

K[1]

S[53] = P13

K[1]

S[6] = P1

K[2]

S[22] = P5

K[2]

S[38] = P9

K[2]

S[54] = P13

K[2]

S[7] = P1

K[3]

S[23] = P5

K[3]

S[39] = P9

K[3]

S[55] = P13

K[3]

S[8] = P2

K[0]

S[24] = P6

K[0]

S[40] = P10

K[0]

S[56] = P14

K[0]

S[9] = P2

K[1]

S[25] = P6

K[1]

S[41] = P10

K[1]

S[57] = P14

K[1]

S[10] = P2

K[2]

S[26] = P6

K[2]

S[42] = P10

K[2]

S[58] = P14

K[2]

S[11] = P2

K[3]

S[27] = P6

K[3]

S[43] = P10

K[3]

S[59] = P14

K[3]

S[12] = P3

K[0]

S[28] = P7

K[0]

S[44] = P11

K[0]

S[60] = P15

K[0]

S[13] = P3

K[1]

S[29] = P7

K[1]

S[45] = P11

K[1]

S[61] = P15

K[1]

S[14] = P3

K[2]

S[30] = P7

K[2]

S[46] = P11

K[2]

S[62] = P15

K[2]

S[15] = P3

K[3]

S[31] = P7

K[3]

S[47] = P11

K[3]

S[63] = P15

K[3]

Fig. 1 Subkey generation technique

Pseudo Code for Encryption Using ATEA 1. 2. 3. 4. 5. 6.

Initialize sum = 0 and delta = 0x9E3779B9 Declare a byte array vv to store the left and right halves of the file data for j = 0 to 16 Initialize i = 0 sum += delta vv[0] += ((((vv[1] > 5) + S[i + 1])) ^ (vv[1] + sum))

430

M. Gupta et al.

Fig. 2 One cycle of encryption using ATEA

vv[1] += ((((vv[0] > 5) + S[i + 3])) ^ (vv[0] + sum)) 8. i = i + 4 9. End for loop 10. Swap left and right halves 7.

Alternate Tiny Encryption Algorithm: A Modified Tiny Encryption …

431

3.4 Decryption It simply involves the inverse operations in reverse order, i.e. 16 cycles of addition and use of the subkeys in the opposite order of encryption. Figure 3 shows an ATEA decryption cycle consisting of two rounds. Pseudo Code for Decryption Using ATEA Initialize delta = 0x9E3779B9 and sum = 16 * delta Declare a byte array vv to store the left and right halves of the file data for j = 0 to 16 Intialize i = 63 vv[1] −= ((((vv[0] > 5) + S[i − 3])) ^ (vv[0] + sum)) 6. vv[0] −= ((((vv[1] > 5) + S[i − 1])) ^ (vv[1] + sum)) 7. sum −= delta 8. i = i − 4 9. End for loop 10. Swap left and right halves 1. 2. 3. 4. 5.

We implemented the proposed ATEA algorithm to make a web application whose architecture is shown in Fig. 4. The application aims to maintain the secrecy of the data files that the client sends to it [13]. Aside from cryptography, the application also supports the upload of encrypted files. This feature allows a user to easily download and install an encrypted file on the server. The files are stored in separate encrypted partitions for each user. Also, the server sends a message to the recipient when the link is sent. Only encrypted files are allowed to be uploaded on the server and links to the uploaded files are automatically generated and returned to the user. Also, the encrypted files uploaded on the server can be easily shared with any other registered users of the application using the share option. The link of the encrypted file on the server is sent as a message to the recipient user which can be easily downloaded by a single click and decrypted if the recipient has the respective key [14].

4 Results ATEA is a simple algorithm to implement. The encrypting and decrypting operations are nearly similar, needing only a key schedule reversal. As a result, the size of the code needed to implement ATEA is nearly halved. The number of rounds in a system depends on a trade-off between efficiency and security. More rounds provide a more security. More rounds, on the other hand, imply inefficiently slow encryption and decryption. Hence, we have reduced the number of rounds to 32 to make these processes fast. And for increasing the security, we

432

Fig. 3 One cycle of decryption using ATEA

M. Gupta et al.

Alternate Tiny Encryption Algorithm: A Modified Tiny Encryption …

433

Fig. 4 Web application architecture

have added the key and subkey generation module which ensures that the subkeys are independently chosen for each round. TEA employs a basic key schedule that breaks down the 128-bit key into four 32-bit chunks and uses them again and again in consecutive rounds which makes it prone to related-key attacks whereas ATEA has a key generated from mouse cursor coordinates which are further converted to a hash key of 128 bits and further divided into 64 unique keys using the proposed subkey generation technique thus making it more random and acting as an added layer of security. XTEA uses a subkey generation technique that provides better security as compared to TEA [9]. The subkeys generation technique of XTEA is simple and easy to crack as compared to our proposed key and subkey generation technique.

434

M. Gupta et al.

We have summarized some similarities and differences between our proposed algorithm ATEA and the existing algorithm TEA and XTEA for encryption and decryption as shown in Table 1. Tables 2 and 3 show the approximate encryption and decryption times of our algorithm and the existing TEA and XTEA security algorithms for a file size of 12 kilobytes with different key sizes. We verified the effectiveness of our algorithm by calculating the Avalanche Effect that refers to the number of modified bits in a ciphertext divided by the number of bits in the ciphertext. An avalanche greater than 50% must always be satisfied by a good cipher [15]. For our proposed algorithm approximately 50% of ciphertext bits differ after every round as shown in Fig. 5. Thus, using ATEA for encryption, the input bits get thoroughly mixed in each round resulting in many changes in output bits making cryptanalysis very difficult. Also, by calculating the average of the avalanche effect in Table 1 Comparison between ATEA, TEA and XTEA algorithms Characteristics

ATEA

TEA

XTEA

Rounds/cycles

32 rounds (16 cycles)

64 rounds (32 cycles)

64 rounds (32 cycles)

Master key size

128 bits

128 bits

128 bits

Key generation

Mouse cursor coordinates

Not unique

Not unique

Key used in each round

Unique

Repeating

Unique

Subkey scheduling algorithm

Yes

No

Yes

Security

Secure and strong

Less secure and weak

Secure and strong

Structure

Feistel cipher

Feistel cipher

Feistel cipher

Performance

Better than TEA

Relatively low

Better than TEA

Table 2 Time taken for encrypting a 12 KB file Key size (in bits)

Encryption time (in ms) TEA

XTEA

ATEA

32

1.173

2.287

1.329

48

1.178

2.572

1.201

64

1.248

2.089

1.092

96

1.208

2.32

1.301

128

1.067

2.301

1.287

160

1.137

2.608

1.053

192

1.39

2.327

1.129

240

1.439

2.866

1.306

Alternate Tiny Encryption Algorithm: A Modified Tiny Encryption …

435

Table 3 Time taken for decrypting a 12 KB file Decryption time (in ms) TEA

XTEA

ATEA

32

1.127

2.249

1.360

48

1.195

2.572

1.239

64

1.241

2.084

1.145

96

1.226

2.299

1.297

128

1.029

2.265

1.301

160

1.093

2.264

1.049

192

1.363

2.278

1.159

240

1.402

2.827

1.401

49.97 42.18

50.5

10

48.43 47.39

58.33 48.43 52.6

9

45.31

48.43 51.9

50

57.28

57.81 53.12

46.87 51.56

46.87 45.31

56.25 53.64

54.16

46.87 49.47

51.03

45.31

50

42.18

Avalanche effect (%)

60

56.25

ATEA 57.81 52.07

TEA 70

50 49.9

Key size (in bits)

40 30 20 10 0 1

2

3

4

5

6

7

8

11

12

13

14

15

16

Number of rounds

Fig. 5 Avalanche effect comparison in ATEA and TEA

TEA which is approximately 49.32% and ATEA approximately 51.77% from Fig. 5, we can conclude that the ATEA shows a better avalanche effect than TEA.

5 Conclusion This paper proposes an improved technique for data encryption, working on the limitations of the previous versions. It shows that there is no significant degradation in the considered cryptographic quality by using ATEA over standard TEA and XTEA.

436

M. Gupta et al.

ATEA provides strong and secure encryption/decryption. Rather than using the primary key directly in the encryption/decryption process for the file data one can use it as a master key to derive unique subkeys and use them for the actual cryptographic processing. Hence ATEA can be considered as a good cryptographic algorithm to be implemented. Although the proposed method offers robust and safe encryption and decryption, it might eventually be improved by increasing performance and decreasing complexity.

References 1. Liu S, Gavrylyako O, Bradford P (2004) Implementing the TEA algorithm on sensors. In: Proceedings of the 42nd annual Southeast regional conference, pp 64–69 2. Sharma M, Arora JB (2017) Cryptography and its desirable properties in terms of different algorithm. IITM J Manag IT 75–81 3. Ramanujam S, Karuppiah M (2011) Designing an algorithm with high avalanche effect. IJCSNS Int J Comput Sci Netw Secur 106–111 4. Narendra K, Pareek VP, Sud K (2010) Block cipher using 1D and 2D chaotic maps. Int J Inf Commun Technol 2(3):244 5. Shoeb M, Gupta VK (2019) A crypt analysis of the tiny encryption algorithm in key generation. Int J Commun Comput Technol 01(01) 6. Wheeler D, Needham R (2004) TEA, a tiny encryption algorithm. In: Proceedings of the fast software encryption: second international workshop. Lecture notes in computer science, vol 1008, pp 363–366 7. Moon MD, Hwang K, Lee W, Lee S, Lim J (2002) Impossible differential cryptanalysis of reduced round XTEA and TEA. In: Daemen J, Rijmen V (eds) FSE 2002. LNCS, vol 2365. Springer, pp 49–60 8. Kelsey J, Schneider B, Wagner D (1997) Related-key cryptanalysis of 3-way, Biham-DES, CAST, DES-X, NewDES, RC2, and TEA. In: Proceedings of the first international conference on information and communication security. Lecture notes in computer science, vol 1334, pp 233–246 9. Lee E, Hong D, Chang D, Hong S, Lim J (2006) A weak key class of XTEA for a relatedkey rectangle attack. In: Nguyen PQ (ed) VIETCRYPT 2006. LNCS, vol 4341. Springer, pp 286–297 10. Hossain MA, Islam MK, Dasand SK, Nashiry MA (2012) Cryptanalyzing of message digest algorithms MD4 and MD5. Int J Crypt Inf Secur (IJCIS) 2(1) 11. Sriramya P, Karthika RA (2015) Providing password security by salted password hashing using Bcrypt algorithm. ARPN J Eng Appl Sci 10(13) 12. Boonkrong S (2012) Security of passwords. Inf Technol J 112–117 13. Benson E, Marcus A, Karger D, Madden S (2010) Sync kit: a persistent client-side database caching toolkit for data intensive websites. In: Proceedings of the 19th International conference on world wide web, USA, Apr 2010 14. Zaman RU, Utkarsh P, Tanwani L, Kumar A (2017) Data encryption and file sharing. Int Res J Eng Technol (IRJET) 04(05) 15. Ramanujam S, Karuppiah M (2011) Designing an algorithm with high Avalanche effect. Int J Comput Sci Netw Secur (IJCSNS) 11(1)

Crystal Clear Analysis of Open–Source Automation Platforms Kiran Jadhav, Mangesh Nikose, and Sagar Shinde

Abstract The advantages of free software are innumerable, and it is highly encouraging to witness many domestic computerization sources that provide free and efficient software program to Internet of Things researchers across the globe. The developers behind such computerization sources have worked hard to create a solid system foundation that anyone can use. Moreover, these sources are freely used by the people to obtain individual responses. Similar to most of the other software systems, for developing an advanced solution, a strong community that is willing to return it to its initial position is required. These characteristics have motivated to compile and compare a greater number of open-source automation software, which is considered as the most exciting domain of IoT. The user can utilize the open source software based on the suitability of the system available and preferred, programming language support for the system, number of users, desired security level, service protocol, and so on. Keywords OpenHAB · Home assistant · Homebridge.io · Jeedom · Eventghost · Pidome · IFTTT · FHEM · Calaos

K. Jadhav (B) Sandip University, Nashik, Maharashtra, India e-mail: [email protected] M. Nikose Department of Electrical and Electronics Engineering, Sandip University, Nashik, Maharashtra, India e-mail: [email protected] S. Shinde PCET’s-NMVP’s Nutan College of Engineering and Research, Pune, Maharashtra, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_33

437

438

K. Jadhav et al.

1 Introduction Home automation software is a software that allows to control and display basic home and office appliances using a PC. Home automation used to be limited to switching the lights and other appliances. However, the possibilities are extensive, allowing customers to create a wireless community, automate Television and Hi-Fi to record both audio and video while the user is away, set up an answering device, and create a climate station by integrating a plethora of different domestic automation technologies into a single advanced technology. According to the UK Department of Trade and Industry [1], a smart home is a place of residence incorporating information and communication technology to connect the key electrical appliances and services that allows them to be remotely controlled, monitored, or accessed. Shareholder network protocols are currently used by many home automation systems. The protocols employed may be exclusive to the business that advanced the machine. The software industry may also prefer this approach because it effectively connects the buyer to their products. However, these can be extensively harmful to the consumer of the home automation device. As a result, it is critical to evaluate a home automation device to ensure that it is built on open protocols.

2 Literature Summary of Open Source Software The proposed study has collected and summarised the available open source software resources in terms of the analysis of most important open-sources of home automation software, including their main features such as system requirements, development language, messaging service, security, speed, and web interface availability, as individual resources and links [2]. The following Table 1 compares some of the most popular open-source home automation software.

2.1 OpenHAB OpenHAB is an open source domestic automation software program available in Java Script. It is set up on premises and integrated to device, which can offer special companies. From 2019, approximately three hundred ties are considered as OSGi modules. Actions, inclusive of latching of lighting are caused through policies, speech instructions or controls at the openHAB client edge. The openHAB assignment started out in 2010 and about 2013, the middle functionality has become a legitimate mission of the Eclipse Foundation under the name eclipse smart home. OpenHAB is primarily based on Eclipse Smart Home and remains as the undertaking for the improvement of bindings. As per Black Duck Open Hub, it remains miles advanced by certainly using one from largest open-source squads present across the globe.

Crystal Clear Analysis of Open–Source Automation Platforms

439

Table 1 Comparison of open source software Open source System software supports

Language Nos. of user

Security

Speed

Service protocol

License

Open-HAB [3]

Java with OSGI, Apache, Karaf

1500

Secure

Very fast

MQTT 2.0

Copyleft

Home Raspberry assistant [4] pi 3

Python 3 (Polymer and YAML)

1000

Secure

Fast

MQTT

Apache 2.0

Jeedom [5]

Hardware hubs and paid support

PHP (Android and IOS)

1000+

Secure

Intermediate MQTT

GPL V2

Io-Broker [6]

ARM with JavaScript 285+ GNU/Linux with node.js and Redis

Secure (Https)

Fast

MQTT and UPnP

MIT

Google Inc

Windows, MAC OS, Linux, Raspberry pi PINE 64

Google Android Node.js, home 5.0+, java, assistant [7] Google app C++, GO

10,000

Highly secure Very fast (AES)

SMS

AGO Control

Raspberry pi wire

YAML

10+

Secure but lot Satisfactory of contradictions

AMQP GNU enterprise public message license

Domoticz [8]

Raspberry pi, UNIX, Apple, Windows

C++

255

Safe

Quite fast

MQTT

GPL V3

FHEM

UNIX, Windows, Fritz Box, Raspberry pi 3

Perl

430

Secure with AES algorithm

Fast

MQTT

GPL V2

Calaos

Wago PLC, C++ Raspberry pi, Cubie Board, Squeezebox

255

Secure

Fast

N.A

GPL V3

Pimatic [9]

Android, IOS

Node.js

N.A

Secure automate HTTPS

Fast enough MQTT

Homebridge Raspberry [10] pi, Linux, IOS, Windows, Docker

Node.js

support class 2 of devices

Highly secure Fast

IFTTT

Node.js, Java

10,000+ Secure

IOS and Android

Fast

GNU GPL

MQTT

Apache 2.0

SMS

Apache 2.0 (continued)

440

K. Jadhav et al.

Table 1 (continued) Open source System software supports

Language Nos. of user

Node-RED

Java 10,000+ Not secure Script and Node.js

Mac, Raspberry pi

Security

Speed

Service protocol

License

Fast

MQTT Apache with TLS 2.0

Fig. 1 A glimpse of OpenHAB

It additionally has a lively consumer era [4]. A glimpse of OpenHAB is shown in Fig. 1. • Merits of OpenHAB [11]: – It can be integrated with more than 200 different technologies, systems and thousands of devices. – Ease of automation. – No cloud required but cloud friendly. – It runs on almost all types of system hardwares.

2.2 Home Assistant Home Assistant is an open source robotization programming, which is intended to be the centralized control framework for smart home gadgets with a focus on nearby

Crystal Clear Analysis of Open–Source Automation Platforms

441

Fig. 2 A glimpse of home assistant

control and security. It can be accessed through an electronic UI, peer applications for Android and iOS, or voice commands for smart devices like Google Assistant and Amazon Alexa. Advanced IoT gadgets, programming applications, and administrations are supported by the measured incorporation segments, which not just incorporate local combinations for nearby network conventions like Bluetooth, MQTT, Zigbee, and Z-Wave, but additionally support for controlling private environments if they give access through a public API for outsider interfaces. When the home assistant programming application is installed on a PC, it serves as a central control framework for home automation. Information from all substances can be used and controlled from scripts trigger mechanizations by utilizing the planning and outline subroutines to control lighting, environment, theatre setups, and apparatus [4]. Figure 2 shows an overview of Home Assistant. • Merits of Home Assistant [12]: – – – –

Works with over 1900 devices Powerful automation All data stays local Home energy management.

2.3 Jeedom Jeedom is ideal for the users searching for a reliable and high-end home computerization package at a reasonable price. Due to its small size and smooth brushed

442

K. Jadhav et al.

metal design, it will fit in any space. It comes with a high-gain Z-Wave+ Antenna, a four-center 1.5 GHZ processor, and an eMMC hard drive. Jeedom is delivered with the Jeedom Service Pack Power programming, which includes administrations that are not available in the open source version. Jeedom’s focal point is only locally viable with the Z-Wave+ (module preinstalled and arranged). If it needs to work with different conventions, the necessary modules and equipment should be downloaded and purchased. Adding the required module implies that each unique home robotization convention or potentially network with different ventures can be coordinated to Jeedom. Jeedom is a hub that empowers a centralization of the entire associated gadgets. Jeedom has a web interface to empower the design of a home mechanization framework and a portable application can also be developed for use through a cell phone or tablet. It is also completely adaptable to whatever requirements of using virtual devices, situations, or modules. Whether remotely, locally, by voice, instant message, or touch screen, you can manage your home whenever and from wherever you want and be alerted whenever an event occurs [5]. Figure 3 depicts a glimpse of Jeedom. • Merits of Jeedom [13]: – Compatible with various protocols – No need to access external servers to work – Flexible and lots of customization settings available.

Fig. 3 A glimpse of Jeedom

Crystal Clear Analysis of Open–Source Automation Platforms

443

2.4 IoBroker IoBroker is open-source software with unique and efficient household components (IoTs), which co-operatively combine into an entire machine (Smart Home System). This gadgets work on their very own but this gives us a complicated data management panel with a graphical user interface, which may be accessed on the neighborhood community with an internet browser. IoBroker is composed of different modules called adapters. Smart devices join IoBroker with adapters. The required adapters can be triggered with one click and further it will be used after a brief setup. IoBroker may be set up on a SoC unmarried-card pc (Raspberry pi, Orange pi, and so on.) or on a computing device PC strolling Windows, Linux. Installing the running device on a Raspberry Pi four USB HDD or SSD is considered as a proper solution. On an old PC, it runs flawlessly under Debian-10, and with java script, one can create automation approaches, connect adapters, and so on. It enables the creation of customized and manipulated strategies. For this reason, there are also Blocky, Node-red, and Scenes; one of the most important adapters is MQTT. Since MQTT allows for establishing a low-cost two-way communication, it is an ideal choice for performing data exchange between a server and a microcontroller [6]. Figure 4 depicts an overview of IoBroker. • Merits of IoBroker [14]: – Unique graphics and attractive interfaces – RUNS ON: Windows, Linux, OSX, Raspberry Pi, ARM or PC.

Fig. 4 A glimpse of IoBroker

444

K. Jadhav et al.

Fig. 5 A glimpse of google home assistant

2.5 Google Home Assistant Google Assistant uses speech instructions, speech based search, and speech based control devices, and removes some of the obligations after user state “OK Google” or “Hey Google” phrases. It is intended to respond user in dialogue exchanges. Google Assistant confers command to smart devices and smart home [7]. Google Home Assistant glimpse is shown in Fig. 5.

2.6 AGO Control AGO control is a framework for tool management. The primary goal is to provide a complete automation solution. It is also used in other applications including agriculture. Previously, manipulate makes use of an AMQP enterprise message bus as a backend and a low footprint protocol, which is readable by users and auto readers, a modern and modular structure, cloud features, and much more. AGO manipulate boasts excellent performance and even works on embedded devices such as Raspberry Pi and multi-plug computers such as Guruplug, Pogoplug, and Sheevaplug. AGO control provides assistance for many peripherals with different protocols, they are 433 MHz trancievers, KNX, Z-Wave, X10, EnOcean, 1wire, Dreambox/Enigma2, Asterisk PBX, Chromoflex USP3 RGB LED dimmer, Onkyoe ISCP AVR, APC Power Distribution Unit, DMX Interfaces by the OLA, Phillips TV units, IRTrans Ethernet purchaser infrared blaster, Webcam guide, Rain8net irrigation controller,

Crystal Clear Analysis of Open–Source Automation Platforms

445

Fig. 6 A glimpse of AGO control

Webcam guide, and BlinkM LED suit. In the past, it was simple to add a list of user suitable tool drivers. Figure 6 shows a glimpse of AGO control.

2.7 Domoticz Domoticz is an absolutely compact home automation device that assist user to supervise and set various devices, like lighting, switches, different devices which sense or measure different parameters like warmth, precipitation, air pressure, radiation, radiation, energy usage/manufacturing, fuel intake, water consumption, etc. Notifications/indicators can also be sent to any mobile tool [8]. A glimpse of Domoticz is shown in Fig. 7.

2.8 FHEM FHEM (TM) is a perl-based GNU-GPL server for developing smart homes. FHEM performs common tasks within the home such as controlling shutters/lamps/heating and recording events such as temperature/humidity/energy consumption. The application runs as a server, and it can be directly controlled via net or phone frontends, telnet, or TCP/IP. To use FHEM, the user must have a server (such as a NAS, RPi, PC, or MacMini) with a perl translator and hardware such as the CUL-, EnOcean-,

446

K. Jadhav et al.

Fig. 7 A glimpse of domoticz

and Z-Wave-USB-Stick to gain access to the sensors. Figure 8 depicts a glimpse of FHEM.

2.9 Calaos Calaos is a powerful open-source domestic automation software program. It functions as a whole set of components to automate house from lighting fixtures to shutter or cameras. It incorporates a complete solution for home automation. The enlightenment foundation libraries are used to strengthen its touch-screen interface. Calaos OS is a whole linux distribution used on its personal. It is primarily based on open-embedded development and provides binary images for a set of machines. Calaos client and server with web-app are mandated for automation. A glimpse of Calaos is shown in Fig. 9.

Crystal Clear Analysis of Open–Source Automation Platforms

Fig. 8 A glimpse of FHEM

Fig. 9 A glimpse of calaos

447

448

K. Jadhav et al.

2.10 Pimatic Pitmatic is a domestic automation framework that runs on Node.Js. It presents a common extensible platform to manage home automation tasks. It defines several schemata for developing exceptional home gadgets and sensors, so that any device can be controlled uniformly and are provided with a commonplace interface. Automation duties can be described with the help of policies in the shape of “if this then that”, where the “this” and “that” element may be completely custom designed via plug-in. More information can be found on the rule of thumb page. The mobile frontend plug-in provides a user-friendly internet frontend with sensor evaluation, device manipulation, and rule definition. Express and jQuery Mobile are being used to build the internet interface. Pitmatic can be extended to various plug-ins, which include functions and integrate existing hardware and software [9]. Figure 10 depicts a glimpse of Pimatic.

2.11 Homebridge.io Homebridge.io is a Node.js based server that acquires less memory and user runs it on a private home community that emulates the iOS HomeKit API. It helps plugin, which are considered as the network-contributed modules that provide a simple bridge from HomeKit to diverse APIs supplied by using producers of smart domestic devices. A glimpse of Homebridge.io is shown in Fig. 11.

2.12 Node Red To connect various edge boards, API, and over the internet facilities in a novel and exciting thought, Node-Red is considered as the ideal solution. It provides a browser to simply utilize a large number of nodes, which can be deployed to its runtime with a single click on. Node-Red includes a browser-based glide editor to make it simple to connect flows by using the palette’s large number of nodes. After the software is triggered, signalling can be configured. JavaScript features may be created within the editor by utilizing a wealthy text editor. A built-in library permits to shop beneficial feature, templates or flows for re-use [10]. A glimpse of Node-Red is shown in Fig. 12.

Crystal Clear Analysis of Open–Source Automation Platforms

449

Fig. 10 A glimpse of pimatic

2.13 IFTTT If This Then That (IFTTT) offers services to control a reaction to different activities inside the smart home framework. IFTTT has integrated with exclusive carrier carriers to deliver timely announcement to IFTTT and follows instructions that execute the replies. In some instances, command interfaces are considered as public APIs. The applications, referred to as applets are simply created graphically. User can create applications and in any other case manipulate IFTTT with an iOS or internet interface or Android application [15]. A glimpse of IFTTT is shown in Fig. 13.

450

Fig. 11 A glimpse of Homebridge.io

Fig. 12 A glimpse of NodeRed

K. Jadhav et al.

Crystal Clear Analysis of Open–Source Automation Platforms

451

Fig. 13 A glimpse of IFTTT

3 Conclusion These are the important open-source home automation software, which can be used by the developers in the field of IoT. The current review comprises a brief and detailed comparative analysis of the most important open-source home automation software by including their main features like system requirements, development language, messaging service, security, speed and web interface availability. The proposed has also analyzed the different licensing techniques in a detailed manner. Some of the open source applications dealt in the report are Home Assistant, openHAB, Jeedom, ioBroker, Google home assistant, Calaos, Domoticz, FHEM, Pimatic, AGO control etc. The user can utilize the perfomred research study to select different entities essential for IoT based automation depending on features and availability of system requirements.

4 Future Work With the proposed comparison on open source IoT based automation and their selections for users, the users can select desired open source software depending upon the system available or vice versa with service protocol for set up communication.

452

K. Jadhav et al.

In further work, there is a scope to set up a few cases with various open source software installed on differ support systems mentioned in Table 1 in order to check results and improve the performance.

References 1. What is a “smart home”?: Smart home energy [Online]. Available: http://smarthomeenergy.co. uk/what-smart-home. Accessed on 10 July 2022 2. Pelaez A (2021) 16 open source home automation platforms to use in 2020. Last accessed 19 Mar 2022. https://ubidots.com/blog/open-source-home-automation/ 3. Baker J (2017) 6 open source home automation tools, build a smarter home with these open source software solutions. Last accessed 14 Dec 2017. https://www.opensource.com/tools/ home-automation/ 4. Saxena S, Jain S, Arora D, Sharma P (2019) Implications of MQTT connectivity protocol for IoT based device automation using home assistant and OpenHab. In: 6th International conference on computing for sustainable global development (INDIACom), pp 475–480 5. Stolojescu-Crisan C, Crisan C, Butunoi BP (2021) An IoT-based smart home automation system. Sensors 21:3784. https://doi.org/10.3390/s21113784 6. Korachentsov A, Vasilevna KA (2019) Automatic control method of indoor lighting using motion sensors in smart home systems. https://doi.org/10.13140/RG.2.2.16213.17122 7. Park MJ, Joshua I (2020) Preliminary study of a google home mini. J Digit For 12(1) 8. Zinca D (2018) experiments with protocols and frameworks used in IoT. In: IEEE 2018 12th international conference on communications (COMM)-Bucharest, 14 June 2018–16 June 2018, pp 277–280. https://doi.org/10.1109/ICComm.2018.8484269 9. Wilhelm S (2021) Activity-monitoring in private households for emergency detection: a survey of common methods and existing disagreeable data sources. https://doi.org/10.5220/001018000 2630272 10. Macheso P, Manda TD, Chisale S, Dzupire N, Mlatho J, Mukanyiligira D (2021) Design of ESP8266 smart home using MQTT and node-RED. In: International conference on artificial intelligence and smart systems (ICAIS), pp 502–505. https://doi.org/10.1109/ICAIS50930. 2021.9396027 11. Why openHAB? [Online]. Available: https://www.openhab.org. Accessed on 10 July 2022 12. Features of home assistant? [Online]. Available: https://www.home-assistant.io/. Accessed on 10 July 2022 13. Jeedom Official Web Page. Available online: https://www.jeedom.com. Accessed on 10 July 2022 14. IoBroker Official Web Page. Available online: https://www.iobroker.net/. Accessed on 10 July 2022 15. S˘al˘agean M, Zinca D (2020) IoT applications based on MQTT protocol. In: 2020 international symposium on electronics and telecommunications (ISETC), pp 1–4. https://doi.org/10.1109/ ISETC50328.2020.9301055

A Review Paper on Network Intrusion Detection System Nongmeikapam Thoiba Singh

and Raman Chadha

Abstract Computer networks are prone to cyber as a consequence of global internet use; as a consequence, academics have developed several Intrusion Detection Systems (IDSs). Identifying intrusions is one of the main significant study topics in data security. It aids in the detection of misuse and attacks as a safeguard for the network’s integrity. Machine Learning, Bayesian-based method, nature-inspired meta-heuristic methods, swarm intelligent approach, and Markov neural network is some ways to find the most effective characteristics and thus improve the effectiveness of Intrusion Detection System (IDS). Over the years, numerous databases have been used to evaluate various projects. This publication provides a comprehensive assessment of IDS with machine learning approaches. Keywords Machine learning · Single classifiers · Hybrid · Ensemble · Misuse identification · IDS

1 Introduction An IDS is used to identify illegal or aberrant conduct. An attack is initiated in a system that is exhibiting unusual activity. Attackers use system flaws, including poor security measures and practices, and program defects like buffer overflow, to cause network breaches. The intruders might be less privileged equipment owners seeking more network access or black hat hackers accused of stealing data from ordinary internet users [1]. Approaches for identifying intruders can focus on identifying exploits or detecting anomalies. Misuse-based IDS monitors network traffic and compares it to a database of signs of known malicious behaviour. Attacks are discovered in the method of anomaly detection when they are matched to activities that depart from typical user activities [2].

N. T. Singh (B) · R. Chadha Department of Computer Science and Engineering, Chandigarh University, Mohali, Punjab, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_34

453

454

N. T. Singh and R. Chadha

NIDS or HIDS are two kinds of IDS. To monitor and assess activity on a given machine, computer network (CN) managers use the HIDS method [3]. When traveling over a system, HIDS offers the advantage of allowing encrypted information to be collected. Since each host must configure and manage data, HIDS has the drawback of being hard to maintain. Additionally, certain Denial-of-Service (DOS) assaults may be able to wipe out HIDS. A network-based intelligently deployed software or HIDS that tracks packets as they transit through the point is called a NIDS. The NIDS consists of two components: one for general monitoring and the other for tracking network conversations. The advantage of NIDS is that it only takes a few people to administer a wide network and is frequently camouflaged from various intruders; products are safe against attack. The disadvantage of NIDS is that it makes it hard to identify an attack during high traffic periods. Many IDS are built on obsolete datasets, such KDD Cup 99, NSL-KDD, and many more, that are devoid of the most recent attack tags. The current initiatives have a poor rate of recognition. This happens as a result of the option to eliminate all unnecessary and irrelevant columns. This occurs when lawful traffic is mistakenly classified as an attack. The FPR expands the IDS’s capabilities while decreasing its effectiveness. The remaining portion of the document is organized as follows. The IDS architecture is contained in Sect. 2. A description of ML algorithms may be found in Sect. 3. Data reduction techniques used in IDS are covered in Sect. 4. Explaining the literature review is Sect. 5. The benchmark datasets for network intrusion detection systems (NIDS) are described in Sect. 6. In Sect. 7, the conclusion is provided.

2 Architecture of IDS The first IDS was created by Dorothy E. Denning in 1986 in collaboration with the SRI International research team. The signature detection element of the double conceptual framework was characterized as containing an attack rule base and an anomaly identification stage using a statistically based analysis to identify novel attacks. IDS became an exciting topic of inquiry in the research community after then. Signature and anomaly IDS are the two most popular forms of IDS. Anomaly-based IDS employs a statistical method to detect actions that differ from typical resource utilization and behavior characteristics. Anomaly-based IDS employed machine learning techniques to create and train IDS models. These models dealt with any incoming suspicious activities. On the other hand, simple statistical methods like mean, median, and quantiles can detect univariate anomalies in datasets. Data visualization and data analysis techniques are also used to detect anomalies. In anomaly-based detection, the proportion of False Positive and False Negative remains high. Figure 1 shows how these IDS work.

A Review Paper on Network Intrusion Detection System

455

Fig. 1 Signature and anomaly intrusion detection system

Following are the various performance metrics used in IDS: True Positive Rate: It is computed as the ratio between the number of accurately predicted attacks and the total number of attacks. We can express TPR as: True Positive Rate =

TP TP + FN

(1)

where TP = True Positive FN = False Negative. False Positive Rate: It is computed as the ratio between the number of normal instances incorrectly predicted as an attack and the total number of normal instances. We can express FPR as:

456

N. T. Singh and R. Chadha

False Positive Rate =

FP FP + TN

(2)

Where FP = False Positive TN = True Negative. False Negative Rate: It indicates when a detector fails to identify an anomaly and classifies it as normal. We can express FNR as: False Negative Rate =

FN FN + TP

(3)

where FN = False Negative TP = True Positive. Accuracy: It may be defined as the proportion of examples that are properly categorized to all occurrences. Accuracy =

TP + TN TP + TN + FP + FN

(4)

where TP = True Positive TN = True Negative FP = False Positive FN = False Negative. Anomaly-based detection procedures must be devised to identify irregular behaviors and achieve optimal accuracy. ML approaches use complex equations. If enough training data is provided and well-trained methods are built, IDS may even be able to detect zero-day assaults. Cyber security managers could use these strategies must make sure security.

3 Overview of ML Approaches ML can be defined as a method of learning algorithms to automatically learn and develop quality characteristics so that they do not need to be specified [2, 4] by utilizing prior experience or example information. The ML model is centered on training sets to forecast distinct class labels per attribute. Machine learning models use existing datasets as model inputs to predict outputs. The machine learning models are fed with sufficient training sets to learn from. The dimensions of the datasets are

A Review Paper on Network Intrusion Detection System

457

Normal Dataset

Data Preprocessing

Feature Selection

Classification Attack

Fig. 2 Block diagram of IDS

reduced. At the same time, the features are extracted. In machine learning, classification is a supervised learning approach where labeled datasets are used in training. The entries in the datasets are labeled with distinct class labels. Supervised Machine Learning models are trained in order to forecast the class labels [5]. Block Diagram of IDS is shown in Fig. 2. ML is often divided into three categories.

3.1 Supervised Learning The samples of the input vector and their corresponding anticipated output vectors make up the database needed for supervised learning training. NB, KNN, ANN, Decision Tree, SVM, Ensemble methods, and Logistic Regression (LR) are a few of the approaches used in this kind of learning [3].

3.2 Unsupervised Learning Unsupervised Learning does not have groups to train from, so it must discover an organization’s information on its own [1]. This is often referred to as self-teaching. Unsupervised methods include the SOM, Apriori method, Éclat approach and outlier detection, Hierarchical Clustering, and Cluster Analysis.

3.3 Reinforcement Learning The system is trained to create a series of decisions using reinforcement learning. The aim is accomplished in a hazy and frequently challenging manner [1]. The system uses trial and error to find a solution to this issue. Several reinforcement learning (RL) techniques include DQN, Q-Learning, SARSA, and DDPG.

458

N. T. Singh and R. Chadha

4 Data Reduction Approaches Used in IDS Due to the enormous size and complexity of information, the weight of ML and data mining (DM) algorithms did not function well with malware detection [4, 6]. These approaches take a long time to categorize assaults, making deployment in realtime situations more problematic. This is due to the increasing amount of features in-network data that Malware Detection must analyze. The quantity and quality of characteristics are important for categorization, as they help us grasp their significance or association [7]. If the number of characteristics chosen is insufficient, classification accuracy will suffer, and if they are in order to be able, generalization will suffer. Studies showed that using feature extraction (FE) methods in IDS improves accuracy and reduces processing costs [5]. FE methods like Principal Component Analysis (PCA), Linear Discriminant Analysis, and Exhaustive Feature Selection help in reducing the number of features in datasets. Maximum effort is made to retain as much information as possible while reducing the number of features. These feature extraction methods also help in handling overfitting, training speed up, and improving data visualization. Lastly, they help in extracting maximum information from the models. As a preprocessing stage, dimensionality and feature reduction algorithms are utilized to enhance quality and reduce attack identification time.

4.1 Feature Selection Feature Selection (FS) strategies are being utilized to find a collection of the great attributes that could enhance the procedure’s ultimate result while generating minimal errors. One objective is to reduce simulation response and storage usage. FS approaches are used in IDS to increase attack detection performance. Principle Component Analysis (PCA), Information Gain and Gain Ratio are among the most used feature selection approaches [8]. Filter and Wrapper are two types of feature selection algorithms that combine various FS methods. • In the Wrapper technique, a classifier serves as a “black box” for assessing optimum properties. Such approaches yield great speculation but may suffer from excessive complexity because to the computational cost of creating the classifier. • Filter approaches don’t use classification to evaluate features and are reasonably resistant to overfitting, but they do so using autonomous estimating methods such as distance measurements, consistency measures, and correlation measures.

A Review Paper on Network Intrusion Detection System

459

4.2 Feature Extraction Showing various samples and columns represent characteristics in a database, with features resulting from both quantitative and subjective discoveries. Feature extraction is a method for reducing the dimensionality of information collecting by choosing a subset so that the accuracy of attack detection accuracy remains unchanged and the amount of time spent discovering is reduced. In the environment, there are a variety of feature extraction approaches. Self-Organizing Maps (SOM), Principle Component Analysis (PCA) and other techniques are instances.

4.3 Clustering Data samples are sorted into data sets in clustering, with information samples in each set being comparable in some manner.

5 Literature Survey An IDS is a system traffic monitoring network that tracks suspicious behavior and transmits out notifications when it is found. It’s a part of the software that investigates a computer or network for malicious activities or policy violations. We’ve compiled a list of IDS that have been explored in this study. Yedukondalu et al. [9] executed ML algorithms to the data set, comparing and evaluating their results. In order to identify intrusion rates, the suggested app uses the SVM and ANN methods. Every technique is utilized to calculate whether the information being sought is allowed or includes any irregularities. While the IDS scans the information requested, if it detects anything malicious material, the request is dropped. These techniques used a feature selection method based on correlation and Chi-Squared to minimize the collection by removing unnecessary data. The preprocessed database is instructed and evaluated with the algorithms to produce notable findings, which improves predictive performance. The testing was performed using the NSL-KDD database. Eventually, an SVM technique achieved 48% accuracy, while the ANN method achieved 97% accuracy. In this sample, the ANN performance was better than the SVM. Gupta et al. [10] a real-time solution for hybrid IDS that employed anomaly detection to find new threats and signature-based identification to identify well-known incursions was proposed. The anomaly identification approach may be used to identify intrusions that escaped the signature-based method, leading to a high proportion of detection value in this investigation. On the last day of the test, the system’s

460

N. T. Singh and R. Chadha

precision achieved a substantial value of 92.65%, and as the approach improved and the machine was taught daily, the percentage of false negatives significantly fell. However, when the model was applied to a big data collection, the issue of a sluggish detection rate surfaced. Lin et al. [11] in both the FFANN and PRANN were used to develop the ANN-based IDS, which used scaled conjugate gradient and Bayesian regularization approaches. Particular result measures were utilized to analyze the work’s quality and competence. The two models have been proved to outperform each other in several output tests on distinct attack observations from the proven effectiveness. Consequently, the FFANN increased precision by 98.0742%. The production’s dependability should be enhanced through testing the design on multiple databases. Hajisalem et al. [12] a single ML classifier was used to create an IDS. They used the NSL-KDD database to test RF and DT methods. The random classifier outperforms the decision tree in accuracy and produces superior outcomes. Both the detection rate and the FPR were not examined in the study. Atefi et al. [13] single classifiers were employed to more accurately identify the assault. The techniques used include SVM, LR, RF, and DT. The NSL-KDD database was used to evaluate the work. The IDS performs best using the RF classifier, according to the study. They also discovered that the RF method had the quickest completion time. The study has the disadvantage of only being able to examine one database efficiently. Ankome et al. [14] proposed for MANET a HCIDM. NS 2.35 was used to create HCIDM. 50 normal MN and five black hole nodes traveled freely in a 4000 m square grid in the experiment. The authors place the black hole on the shortest path between the source and DN. The HCIDM’s productivity is assessed using measures such as PDR, packet loss, and throughput. The system with HCIDM had average network congestion of 26.92, but CCIDM had an average of 19.85, which could be due to HCIDM’s maximum bandwidth. Throughout a BHA, the average total performance of the system with HCIDM is 23.23 Mbps, a 94% increase. Nie et al. [15] GAN was used to create a new intrusion detection tool (Generative Adversarial Network). For feature learning, the GAN-based technique may extract less-dimensional characteristics from actual system flows. MATLAB was used to develop the suggested GAN-based system. The proposed approach was evaluated using the CIC-DDoS2019 and CSE-CIC-IDS2018 databases. Their strategy can significantly enhance intrusion detection accuracy based on the simulation results. The accuracy of both databases was 95.32% for the CSE-CIC-IDS2018 dataset and 98.53% for the CIC-DDOS2019 dataset. Sankaranarayanan et al. [16] the intruders are effectively identified using the suggested Secure Intrusion Detection approach, which uses the RSA approach. The simulation is run on Ubuntu and the Network Simulator NS2.34 environment. The findings showed that in the existence of intruders, the suggested secure IDS approach provides a higher PDR. Even when routing overhead is on the higher side, it is permissible when authors need to detect attackers.

A Review Paper on Network Intrusion Detection System

461

Huang et al. [17] based on traffic pattern learning, created an effective IDS for VSNs. In order to characterise the dynamic properties of network traffic in Visual Sensor Networks, the suggested approach builds a traffic model. By utilising Omnet++ to create a VSN, it is possible to compare the outcomes of our IDS with the traffic data that has been produced in order to find the best feature set for traffic pattern learning. An HSOM is then used to comprehend traffic circumstances and identify intrusions. In order to hasten the HSOM’s training and improve its understanding of attack patterns, an active learning approach is also developed. The suggested approach provides excellent real-time performance and great detection precision.

6 NIDS Datasets KDD CUP’99: It is a popular NIDS dataset with 5 million training and 2 million testing records. Each record includes 41 features. The attacks are labeled as DoS, R2L, U2R, and Probe [18]. UNSW-NB15: The Australian Center for Cyber Security releases this NIDS dataset. It contains around 2 million records with 49 features. This dataset includes nine attack types: Shellcode, Worms, Generic, Reconnaissance, Port Scans, DoS, Fuzzers, and Backdoors [19]. NSL-KDD: This is the updated version of the KDD Cup’99 dataset. It contains four different attack types, including DoS, R2L, and Probe, and 41 characteristics [20]. CIC-IDS2017: This NIDS dataset is made available by the Canadian Institute of Cyber Security. It contains normal flows as well as updated real-world attacks. Brute force, botnet, HeartBleed, DoS, web, DDoS, and infiltration are some of the attack techniques it may use [21]. Kyoto 2006+: Kyoto University produced this dataset from network traffic logs. It has exactly 24 statistical characteristics [22]. CSE-CIC-IDS2018: This dataset contains seven attack types: Web, Infiltration, Brute Force, DDoS, Botnet, DoS, and Heart Bleed [23].

7 Conclusion The development of ML introduces new techniques to IDS, with many researchers and academics using various types in the construction of IDS designs. In today’s environment, intrusion detection is crucial for information security, so ML-based apps have aided in the discovery of new assaults. In recent years, several classifiers, such as hybrid systems and ensemble learning approaches, have dramatically enhanced

462

N. T. Singh and R. Chadha

the reliability of attack detection strategies. However, there is still a problem with the proportion of FP and FN. Authors encourage scientists and researchers to consider using more approaches with a higher accuracy rate.

References 1. Sangkatsanee P, Wattanapongsakorn N, Charnsripinyo C (2011) Practical real-time intrusion detection using machine learning approaches. Comput Commun 34(18):2227–2235 2. Nader P, Honeine P, Beauseroy P (2016) Detection of cyber attacks in a water distribution system using machine learning techniques. In: 6th international conference on digital information processing and communications, ICDIPC 2016 3. Jabbar MA, Aluvalu R, Reddy SSS (2017) Cluster-based ensemble classification for intrusion detection system. In: ACM international conference proceeding series, vol Part F1283, pp 253–257 4. Verma P, Anwar S, Khan S, Mane SB (2018) Network intrusion detection using clustering and gradient boosting. In: 9th international conference on computing, communication and networking technologies (ICCCNT), pp 1–7 5. Shakya S (2021) Modified gray wolf feature selection and machine learning classification for wireless sensor network intrusion detection. IRO J Sustain Wirel Syst 3(2):118–127 6. Denning DE (1986) An intrusion-detection model. In: IEEE symposium on security and privacy, pp 118–131 7. Saha S, Sairam AS, Yadav A (2012) Genetic algorithm combined with support vector machine for building an intrusion detection system. In: Proceedings of the international conference on advances in computing, communications and informatics. ACM, pp 566–572 8. Vergara JR, Estevez PA (2014) A review of feature selection methods based on mutual information. Neural Comput Appl 24(1):175–186 9. Yedukondalu G, Hima Bindu G, Pavan J, Venkatesh G, Sai Teja A (2021) Intrusion detection system framework using machine learning. In: Third international conference on inventive research in computing applications (ICIRCA) 10. Gupta S (2016) Analyzing the machine learning algorithms—Naïve Bayes, random tree, and support vector machines SVM using the kdd99 data set to predict and classify the, vol 2, pp 452–459 11. Lin WC, Ke SW, Tsai CF (2015) CANN: an intrusion detection system based on combining cluster centers and nearest neighbors. Knowl Based Syst 78(1):13–21 12. Hajisalem V, Babaie S (2018) A hybrid intrusion detection system based on ABC-AFS algorithm for misuse and anomaly detection. Comput Netw 136:37–50 13. Atefi K, Yahya S, Rezaei A, Hashim SHBM (2016) Anomaly detection based on profile signature in network using machine learning technique. In: Proceedings—2016 IEEE region 10 symposium, TENSYMP 2016, pp 71–76 14. Ankome T, LusilaoZodi G-A (2021) Hierarchical cooperative intrusion detection method for MANETs (HCIDM). In: 15th international conference on ubiquitous information management and communication (IMCOM) 15. Nie L, Wu Y, Wang X, Guo L, Wang G, Gao X, Li S (2021) Intrusion detection for secure social internet of things based on collaborative edge computing: a generative adversarial networkbased approach. IEEE Trans Comput Soc Syst 1–12 16. Sankaranarayanan S, Murugaboopathi G (2017) Secure intrusion detection system in mobile ad hoc networks using RSA algorithm. In: Second international conference on recent trends and challenges in computational models (ICRTCCM) 17. Huang K, Zhang Q, Zhou C, Xiong N, Qin Y (2017) An efficient intrusion detection approach for visual sensor networks based on traffic pattern learning. IEEE Trans Syst Man Cybernet Syst 47(10):2704–2713

A Review Paper on Network Intrusion Detection System

463

18. Bay S (1999) The UCI KDD archive. University of California, Department of Computer Science, Irvine, CA. http://kdd.ics.uci.edu 19. Moustafa N, Slay J (2016) The evaluation of network anomaly detection systems: statistical analysis of the UNSW-NB15 data set and the comparison with the KDD99 data set. Inf Sec J A Global Perspect 25(1–3):18–31. https://doi.org/10.1080/19393555.2015.1125974 20. Tavallaee M, Bagheri E, Lu W, Ghorbani AA (2009) A detailed analysis of the KDD CUP 99 data set. Paper presented at: proceedings of the IEEE symposium on computational intelligence for security and defense applications. IEEE, Ottawa, ON, Canada, pp 1–6 21. Abdulhammed R, Musafer H, Alessa A, Faezipour M, Abuzneid A (2019) Features dimensionality reduction approaches for machine learning based network intrusion detection. Electronics 8(3):322. https://doi.org/10.3390/electronics8030322 22. Song J, Takakura H, Okabe Y, Eto M, Inoue D, Nakao K (2011) Statistical analysis of honeypot data and building of Kyoto 2006+ dataset for NIDS evaluation. Paper presented at: proceedings of the 1st workshop on building analysis datasets and gathering experience returns for security. Salzburg, Austria, pp 29–36 23. Sharafaldin I, Lashkari AH, Ghorbani AA (2018) Toward generating a new intrusion detection dataset and intrusion traffic characterization. Paper presented at: proceedings of the 4th international conference on information systems security and privacy (ICISSP). Madeira, Portugal, pp 108–116

ESP32 Based Irrigation System M. Koteswara Rao, M. Satish Kumar, M. Jaijaivenkataramana, and Ch. Sai Sowjanya

Abstract In India, Agriculture is remaining as the major occupation among people. Farmers cultivate various crops depending on the type of land and season. Additionally, farmers also irrigate agricultural land while cultivating. However, farmers cannot predict how much water they will use for irrigation. Currently, farmers are supplying water to the field without knowing the moisture content of the soil. If this continues, there will be a scarcity of water, and underground water will also be depleted and may not be available for future generations. Furthermore, the crops are destroyed if more water is supplied than the required amount. This research study has utilized an advanced technology called Internet of Things (IoT) to design and develop agricultural monitoring systems for evaluating soil moisture content and other agricultural parameters. This system includes soil moisture sensors for measuring the soil moisture content. Rain drop sensor, water level sensor, and DHT11 sensor are used for measuring rain, field water level content, temperature and humidity. The outputs from the sensors are sent to the ESP32 module, which then sends it to the motor (used for irrigation and ejection). These values can also be visualized in the ThingSpeak cloud platform. Keywords NodeMCUESP-32 · Soil moisture · Raindrop · Water level · DHT11 sensors · Cloud

1 Introduction Agriculture is important in our country because it highly contributes to India’s GDP. Agriculture is also important to our country’s population. However, with the increased use of synthetic fertilizers, soil quality is deteriorating. Furthermore, by not knowing M. Koteswara Rao (B) · M. Satish Kumar · M. Jaijaivenkataramana · Ch. Sai Sowjanya Department of Electrical Communication Engineering, Sri Vasavi Engineering College and Pharmacy (Autonomous), JNTUK, Tadepalligudem, India e-mail: [email protected] M. Satish Kumar e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_35

465

466

M. Koteswara Rao et al.

how much water must be used for irrigation, the water supply is depleted resulting in long-term consequences for future generations. Each agricultural field has a different irrigation value. Furthermore, the moisture content of soil varies from place to place. Since irrigation is dependent on soil moisture content, farmers can easily determine how much water to use by gaining an adequate knowledge on the moisture content of the soil [1, 2]. In addition, water is saved. In this case, a raindrop sensor is used to detect the raindrops, DHT11 sensor measures the temperature and humidity on that particular day, and the water level sensor detects the level of water. These inputs are then sent to the ESP32 board. Finally, the output is sent to the motor, and then the data gets saved in the ThingSpeak cloud [3, 4].

2 Literature Survey Even though agriculture plays a major role in contributing to the country’s economy, farmers are still using the traditional farming methods. Irrigation challenges result in the wastage of underground water and water bodies. To eliminate the problems associated with water wastage, Das et al. [1] proposed an innovative model to avoid water wastage in the process of irrigation by automating the water control process. Vadapalli et al. [2] have utilized the cloud platform known as Thingspeak to collect data on the field parameters of an automatic motor pumping system. As water level systems has become necessary for farmers at any climatic conditions, Suresh et al. [5] have briefly described about the incorporation of new technologies in handling the ever-changing weather conditions. Further, Babiuch et al. [4] have described about the working principle and working environment of ESP32.

3 Proposed System The proposed system has used ESP32 and some sensors like soil moisture sensor, raindrop sensor, DHT11 sensor and water level sensor. The sensor will then be sent to the cloud by using an inbuilt Wi-Fi module for ESP32-WROOM-32.

3.1 ThingSpeak ThingSpeak is an IoT platform as in Fig. 1; it is a modern cloud platform with a user-friendly interface, which allows to store the sensor data, visualize the data, and then analyze it by using MATLAB or any other software. This cloud application works with HTTP and MQTT IoT protocols.

ESP32 Based Irrigation System

467

Fig. 1 ThingSpeak cloud platform

3.2 Soil Moisture Sensor The soil moisture sensor is designed specifically to measure the moisture content of the soil. The soil moisture sensor is highly used in a variety of applications, including root zone measurements and regulating the existing conventional irrigation timer. Soil senor values are shown in Fig. 2.

Fig. 2 Soil sensor values

468

M. Koteswara Rao et al.

Fig. 3 Raindrop sensor values

3.3 Raindrop Sensor The raindrop sensor is a device that detects raindrops. It consists of two modules: a rain board that determines the rain droplets and a control module that differentiates analog values and converts them to digital values. Raindrop sensors are used in a variety of applications, including home automation and protecting the internal parts of an automobile from rain [5]. Raindrop sensor values are shown in Fig. 3.

3.4 DHT11 Sensor The DHT11 is a temperature and humidity sensor with a dedicated NTC for monitoring the temperature and an 8-bit microcontroller for monitoring the temperature and humidity output values. This sensor is used to measure the humidity in medical equipment, home automation systems, automotive, and other weather control applications as in Fig. 4.

3.5 Water Level Sensor The water level sensor module has a series of parallely exposed lines, which measure droplets or water volume to determine the water level as I Fig. 5 [6]. Water level sensor used in different applications to measure water levels in tanks help to control water

ESP32 Based Irrigation System

469

Fig. 4 DHT11 sensor

levels. This can also be used to automatically turn ON/OFF pumps and life station switches.

3.6 Methodology of the Proposed Work The overall block diagram of the proposed work is depicted in Fig. 7. Each and every block has a detailed explanation of each sensor included in the proposed work. The block diagram consists of various sensors like soil moisture sensor, raindrop sensor, water level sensor, and DHT11 sensor. These sensors are then connected to the ESP32 board module. Here, the input values are given and the output values are fed to the actuators and finally the data is stored in cloud platform i.e., Thingspeak [7]. Soil moisture sensors, rain drop sensors, water level sensors, and DHT11 sensors are some of the sensors used in agriculture. Analog sensors include the soil moisture sensor, rain drop sensor, and water level sensor. The DHT11 sensor is a digital sensor. The ESP32 module includes a Bluetooth and Wi-Fi module. The sensors are connected to the ESP32 module in this case. Both analogue and digital values can

470

M. Koteswara Rao et al.

Fig. 5 Water level sensor values

be accepted by ESP32. The output is routed to actuators for additional processing. Here, the rain drop sensor is used to sense the rainfall and the sensed data is then provided to the actuators for performing automatic irrigation in field [8]. Moreover, during rain fall the underground water usage gets decreased. Here, the water level sensor is used for sensing the water level of the field. Here, the water levels differ based on the different types of agricultural field. If the water level exceeds the threshold value, the water is moved to nearby water bodies for future use. The DHT11 sensor is used for measuring the humidity and temperature of the agricultural field. DHT11 sensor consists of a capacitive humidity sensing element and a thermistor for sensing the temperature. Here, the cloud platform used here is Thingspeak. Every sensor and actuator value is uploaded to the cloud platform Thingspeak. All sensor values are shown in Fig. 6.

4 Proposed System See Fig. 7.

4.1 Methodology of the Proposed Work The process flow diagram is shown in Fig. 8.

ESP32 Based Irrigation System

471

Fig. 6 All sensor values

The proposed research focuses on the interfacing of sensors such as soil moisture sensors and rain drop sensors. The motor will turn ON when the moisture content of the land rises, otherwise it will turn OFF. If there is rain, the motor will be turned OFF; otherwise, it will be turned ON. If rain falls but the required threshold value of moisture content is not present in the soil, the motor will turn ON. The two conditions will be checked here. A water level sensor is also used to monitor the water levels in the field. If the level of water exceeds required level, the water will be removed by using an ejection motor. DHT11 sensor, a digital sensor will be used to measure the temperature and humidity of the agricultural field. All the sensors are linked to the ESP32 microcontroller [9, 10]. The measured parameter values will be displayed

472

M. Koteswara Rao et al.

Fig. 7 Block diagram

and saved in the Thingspeak cloud platform. The project’s hardware configuration is shown in Fig. 9. Sensor values on serial monitor are shown in Fig. 10.

5 Conclusion The primary goal of this ESP32-based irrigation system is to be more innovative, userfriendly, resourceful, and efficient than the existing systems. Different agricultural parameters such as soil moisture, temperature, humidity, water level, and raindrop sensing are continuously measured. The data for these parameters is then uploaded to the cloud. By using the proposed system, the farmer can learn about the field’s current parameters at any time and from any location by using these updated values.

ESP32 Based Irrigation System Fig. 8 Flow chart

473

474

M. Koteswara Rao et al.

Fig. 9 Hardware setup of ESP32 irrigation based system

Fig. 10 Sensor values on serial monitor

References 1. Das K, Sahu S (2020) Smart farming using IoT. Int Organ Sci Res 15(02) 2. Vadapalli A, Peravali S, Dada VR (2020) Smart agriculture system using IoT technology. Int J Adv Res Sci Eng 09(09) 3. Anitha A, Sampath N, Jerlin MA (2021) Smart irrigation system using internet of things. In: International conference on emerging trends in information technology and engineering 4. Babiuch M, Foltynek P, Smutny P (2020) Using the Esp32 microcontroller for data processing. In: International Carpathian control conference 5. Suresh N, Hashiyana V, Kulula VP, Thotappa S (2019) Smart water level monitoring system for farmers. In: International conference on emerging trends in information technology and engineering 6. Sharma A (2020) Review on IoT based water level sensing and controlling. Int J Eng Res Technol 09(07) 7. Alsahi QN, Marhoon AF (2020) Design health care system using Raspberry pi and Esp32. Int J Comput Appl 36 8. Krishna BV, Priyanka K (2014) Soil moisture sensor design for crop management system by using cellular communication. Int J Adv Res Electr Electron Instrum Eng 03(10) 9. Sai Ram KS, Gupta ANPS (2021) IoT based data logger system for weather monitoring using wireless sensor networks. Int J Eng Trends Technol 32 10. Tyagi A, Gupta N, Navani JP, Tiwari R, Gupta A (2017) Smart irrigation system. Int J Innov Res Sci Technol 03(10)

RFID (Radio Frequency Identification) Tag Collision Risk Mitigation Analysis and Avoidance Aditya Sukhwal, Gourab Kundu, and Chandrani Chakravorty

Abstract RFID (Radio Frequency Identification) transponder is a small beacon tag responsible for transmitting the radio waves in the range of 30 kHz to 3 GHz to an antenna. The antenna is then connected to the reader, which sends and receives signals from the antenna. The use of multiple tags in a perimeter increases the chance of tag collision. Tag collision occurs when multiple tags send signal to the reader at the same time. Such a situation can lead to miscommunication between tags and readers present within the same perimeter; such miscommunication between reader and tags can lead to the failure of the entire RFID system. To avoid such a situation, an algorithm mechanism helps to avoid collision risks. The main goal is to collect data and identify various ranges and proximity in which an RFID tag collision may occur, as well as to conduct preventative analysis to avoid such failures. Keywords Tag · Transponder · Reader · Antenna · Collision

1 Introduction Radio Frequency Identification (RFID) tag is the small beacon device, which emits the radio wave signals. The signals emitted by the RFID tags are received by an antenna connected to the main device, which is called as reader. Reader is considered as the main brain of RFID tag [1] as it is responsible for converting the signals into raw data, which then gets stored in a database and from there all the tags database can be maintained and manipulated. The reader communicates to the antenna to obtain the radio waves signal emitted from the tags [2]. A. Sukhwal (B) · G. Kundu · C. Chakravorty Department of MCA, RV College of Engineering, Bengaluru, India e-mail: [email protected] G. Kundu e-mail: [email protected] C. Chakravorty e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_36

475

476

A. Sukhwal et al.

Fig. 1 Radio frequency identification mechanism

In a specific area, the tags have a fixed perimeter. When a tag leaves the perimeter, the antenna stops receiving signals from the tag, and thus a security alarm can be triggered, which is essentially a use case of RFID systems in supply chain management, where it is easy to keep track of inventory and avoid the risk of losing inventory [3]. One common example for RFID use case is the attachment of RFID tags to the clothes at cloth store, these tags are attached to clothes so until the cloth is purchased the tag is not removed, hence if anyone take the cloth out of store without buying the alarm will set ON and burglar will get caught. RFID tags have unique identity to recognize automatically. Additionally, RFID tags are tough and waterproof as they still work after getting dropped into water. Figure 1 shows the working mechanism of a RFID system. Here, a transponder tag emits the radio wave signal in a perimeter, wherein the signals emitted by the transponder gets received by an antenna, which is further connected to the main brain of RFID system called RFID Reader. The RFID reader gets connected to a computer and also to a local switcher throughout which multiple reader will be connected to a centralized cloud based computer. The central connection of RFID reader ecosystem expands the area of reader especially when a system is required for large warehouses. Such system provides an edge in terms of connecting an ecosystem to the cloud based distributed system, which can be used to trace the inventory from any place in the world via internet. Antennas are responsible for transmitting the electromagnetic waves.

2 Types of RFID Tags There are mainly 3 types of RFID tags, all 3 types of tags have different applications and use cases, as shown in Fig. 2. The classification of different types of tags have their variation in specifications. All of them provide different radio waves and wavelength ranges and hence the cost of operating them is also directly proportional to their specification (Table 1). A. Low Frequency transponder is highly cost effective and operates at the range of 30.0–300.0 kHz. The absence of battery source makes them long lasting with less maintenance cost. Application of these type of tags are electronic identity card, bus passes etc.

RFID (Radio Frequency Identification) Tag Collision Risk Mitigation …

477

Fig. 2 Two type of collision present in RFID. a Reader and tag. b Reader and reader [4]

Table 1 Comparison of 3 types of RFID transponders Types of RFID tags Low-frequency (RFID)

High-frequency (HF)

Ultra-high-frequency (UHF)

Frequency range

30–300 kHz

3–30 MHz

300 MHz to 3 GHz

Common frequency

125 kHz or 134 kHz

13.56 MHz (NFC)

860–960 MHz (UHF Gen2)

Relative cost

$$

$$–$$$

$

Read range

≤10 cm

≤30 cm

≤100 m

Benefits

More resistant to Higher memory Lower cost, with good interference by liquids capabilities, NFC tags read range and fast and metals can function as both read rates reader and tag

Common applications Animal tracking, automobile inventorying

Promotional packaging and labels, contactless payment, library collections

Inventory control, item-level tracking, supply chain visibility and efficiency

Credit Resource label

B. High Frequency transponders are the cost-efficient NFC (Near Field Communication) tags. NFC used here is mobile phones, which helps to transfer data at a lightning speed by including more memory. It works as both integrated reader and transponder at the same time and cover a range up to 30 cm. C. Ultra-High-Frequency transponders have a battery, which allows it to emit a wide range of radio signals, such as 860.0–960 MHz and a perimeter range of 100 m. UHF transponders are used in the supply chain management applications

478

A. Sukhwal et al.

used for inventories. These tags can be seen in shopping malls connected to garments and apparels to prevent theft. They have a high operating range of up to 100 m. D. Active RFID Tags includes a small battery, which power them for carrying out active control transmission of radio waves with antenna, meanwhile Passive RFID tags get energized when it comes into proximity of radio waves emitted by the antenna.

3 Background Types of tag collisions which occur in highly dense environment are usually classified into two types, Reader and Tag collision and Reader and Reader collision. The first type, reader and tag collision occurs when one tag emit signals to two readers at the same time. The second type, reader and reader collision occurs when a tag is supposed to come under reader X enters the proximity of reader [4]. ALOHA algorithm and binary search algorithms can be classified as the two of major algorithms which are specifically used to avoid the RFID tag collisions. At present, the algorithms that are used to avoid the RFID tag collisions are dependent on TDMA (Time Division Multiple Access) [5]. Different types of RFID tags operate at different frequencies and to optimize the tags for high frequency the inclusion of the battery is needed in High Frequency (HF) and Ultra High Frequency (UHF) based RFID systems. By implementing the two way handshake mechanism, RFID systems assist in avoiding different cyber-attacks [1, 6]. Q algorithm has been massively used by RFID systems since 2005 due to its tag sorting and handling capabilities, Q algorithm collects the responses from the tags and group them according to their response time and with help of various iterations it identifies the tags [7]. The compact exclusion validation method avoids limiting tags by testing them one by one in each slot. The inclusion of a key filtering method allows to perform fast tag filtering while sorting the necessary tags [8]. Multiple readers are frequently present in large warehouses and inventory tracking environments; as the number of readers increases, so does the number of tags. For such a complex system, it is critical that all readers remain critically coordinated in real time. Communication latency is considered as a major issue in such sophisticated systems [9]. Dynamic and backtracking based binary search algorithm is used to prevent tag collisions and make the identification process more time efficient. The binary search algorithm outperforms other algorithms in identifying tags [10].

RFID (Radio Frequency Identification) Tag Collision Risk Mitigation …

479

Tag collision can be avoided by implementing and modifying multiple existing protocols that are fundamentally based on ALOHA, Tree, and other hybrid protocols. The complexity and cost of the RFID network are two components that are primarily concerned with the entire RFID system setup, where the efficiency of the entire ecosystem is directly related to the cost and complexity of the RFID network [11]. When an RFID tag does not respond to a reader within a specified time frame and returns a null response, the reader terminates the idle slots. This is where the M optimal algorithm comes into effect; it is a splitting algorithm in which the value of M is carefully selected when the value of M is greater than 2 [12]. The DFSA algorithm has the capability to recover almost 99% of communication channels present in RFID system. With the help of DFSA reader, the algorithm implements a Field Programmable Gate Array (FPGA) and then the entire RFID ecosystem gets verification through the communication tests accompanied by the tags used in commercial spaces [13]. There are many fields where RFID is being used in today’s world. The applications of RFID have been significantly increasing since the inception of RFID system. The RFID system admin and any unauthorized user can use RFID card to gain access to confidential places. This limits the application of RFID tags in high intensity defense and security areas [2]. Passive RFID tags emit radio wave signals in a proximity, which further acquired by the antenna. By using the amount of energy emitted, the distance of the tag from RFID system can be measured. After receiving the signals, the reader goes to the idle state. Here, the overall aim is to increase the efficiency of slots by incorporating K-means algorithm [3]. Ultra-High Frequency RFID transponders may have a range up to 50 m, at the same time it results in high power consumption to acquire such high range radio waves. Bayesian algorithm with filter results in the transmission of power in UHF based RFID tags. The main intent of this study is to save the power when RFID transponder remains in idle states by using machine learning algorithm [14]. RFID systems are often vulnerable to relay attacks, which are referred as manin-the-middle attacks in which the communication between sender and receiver is disrupted by a third-party unauthorized user, and such interception poses a significant risk to RFID systems. Lightweight method provides the best solution for reducing the likelihood of such successful relay attacks [15]. Existing RFID system challenges are the actual reason why it still stands far from the market capital across the globe. According to largest RFID manufacturers such as Zebra, Alien Technology Inc., and Tyco Retail Solutions, RFID’s future sustainability is dependent on how efficient the existing research and development proves to be in the future. RFID can also be used to track the emerging needs of humankind such as vaccines and medications in times of pandemic [16, 17]. Table 3 demonstrates the comparison among different anti-collision algorithms with severity of failure. It also mentions the complexity level of different algorithms along with the application of these algorithms in suitable domain along with its cost and efficiency.

480

A. Sukhwal et al.

4 Types of Tag Collision Tag collision is mainly classified into two different categories: Reader and Tag collision (Fig. 2a), Reader and Reader Collision (Fig. 2b). Whenever two readers present in same vicinity and send the messaging signal to the tag located in their perimeter range then there are high chances that the tag won’t be responding to any of them. In exceptional circumstances, if the frequencies of the readers are dissimilar, the tag will not respond to any of the readers. In such a Reader to Tag Collision, the tag will never listen to the communication requests send by the readers [14]. In some cases, when a reader send a strong signal even when the reader’s interrogation perimeter is small, the reader may develop a large interference range. Reader and Reader Collision occurs when multiple readers are placed in the same interference range and use the same frequencies of other readers at the same time, as shown in Fig. 1 (Fig. 2b) [4, 5].

5 Risks of Tag Collision Tag collision is referred as a major threat to any of the RFID system, especially when it comes to the real-life application, the collision of tag leads to more complexity in supply chain management system. Hence, it is important to detect such situation and handle it at the earliest. Such problem occurs in the presence of a huge number of Tag/Transponder devices in a perimeter of the RFID system. The tags gets energized by the multiple RFID systems and starts reflecting the radio wave signals back to the multiple or a single reader simultaneously [15]. I. II. III. IV.

Failure of the RFID system Losing the inventory item tracking details Security breach in the vicinity Theft detection if the valuables are attached to RFID tag.

6 Avoiding RFID Tag Collision Failure of an RFID system in a supply chain management challenge poses a significant threat to inventory management; such failures can temporarily suspend the supply chain of carriers and goods; therefore, having a sustainable and robust RFID system should be given utmost importance. RFID systems can be made more secured by using cryptographic algorithms and two-way handshakes. Looking at all of the challenges, tag collision remains as one of the most recurring and difficult challenges that RFID systems face in today’s world. Since its inception, various algorithms such as DBTSA and Q protocol have proven to be more efficient than previous algorithms. The below-mentioned are the description of two of the most versatile and widely

RFID (Radio Frequency Identification) Tag Collision Risk Mitigation …

481

famous anti tag collision algorithms, which has proven to be highly effective against tag collision.

6.1 Dynamic Binary Tree Slot ALOHA (DBTSA) DBTSA (Dynamic Binary Tree Slot ALOHA) [12] was discovered as a leading constructive anti-collision tag algorithm. DBTSA employs dynamic frame utilization and efficient split frame in the algorithm to improve the effectiveness of high frequency passive tag anti-collision performance. The proposed algorithm presented in Fig. 3 provides a non-estimation based protocol, which maintains the protocol effectiveness even as the number of transponders in the RFID system increase. To achieve such a high efficiency, the algorithm focuses on reducing the idle slots and time required for tag identification [12]. As shown in Table 2 the DBTSA is more focused on adjusting the frame dynamically when the efficiency is increased by reducing the idle slots and identifying the tags in minimal response time. Here, the DBTSA achieves an efficiency of 0.441, which is 0.21 times higher than the traditional DFA-OS. DBTSA uses dynamic frame allocation by having a non-static frame size, which can store a higher or smaller number of tags depending on the Q-value.

Fig. 3 Dynamic frame adjustment algorithm (DBTSA) [12]

482

A. Sukhwal et al.

Table 2 DBTSA efficiency matrix [12] Proposed algorithm—dynamic binary tree slotted ALOHA (DBTSA) Number of tags

Number of idle slots

Number of collisions

Identification time (ms)

Number of iterations

System efficiency

0–50

31–38

30–34

140–180

80–90

0.441–0.420

100–500

392–480

344–345

800–970

850–1050

0.411–0.400

Table 3 Comparison of ALOHA based tag anti-collision protocol Protocol name

PA

SA

FSA

DFSA

Q

Protocol feature

Tags send signals at random time intervals, in case of collision tags resends the signal to reader

Tag sends ID in synchronous form of time slots, if collision then resends in random time interval

All tags respond once per frame only

Tags send signal once every window. Reader set tag analysis function to change frame size

Reader sets value of Q dynamically based on types of reply sent by tags

Disadvantages

When number of tags is high in vicinity the collision increases

High quantity of Uses fix tags results in size of more collision frame so reader requires synchronization with tag

Can’t process to next frame without finishing the current frame process

May face lower throughput while adjusting value of Q when frame size is greater than quantity of tags

Efficiency (%)

18.40

36.80

36.80

42.60

36.80

System cost

Very low

Low

Expensive

Expensive

Expensive

Complexity

Very easy

Easy

Medium

High

Medium

6.2 Q Protocol Q protocol remains as a basic framework for Dynamic Frame Slotted ALOHA (DFSA), which centers on manipulating the size of the frame. Based on the last feedback received, RFID tag sends the value of Q to tag. Depending on number of tags considered per round, QNew is calculated. Q algorithms have two fundamental variables Q and a constant remaining C, where the range of the variable Q can be from 0.0 to 15.0 (in Fig. 4). The Q protocol uses 3 different types of commands: • Query Command: It is transferred through reader to the transponder present in the vicinity and forces the transponder to choose a slot number to initialize the process of identification with providing a new value assigned to the Q.

RFID (Radio Frequency Identification) Tag Collision Risk Mitigation …

483

Fig. 4 Q protocol flow chart [11]

• Query Adjust Command: It is used to send instructions to every transponder to decrease, grow and maintain the value of Q unchanged. According to the algorithm value of Q, it can be increased or decreased based on the value of C. • Query Rep comes to the picture when the tags are supposed to decrease the value of Q. DBTSA (Dynamic Binary Tree Slot ALOHA) and Q algorithm both have their own pros and cons, DBTSA is proven to provide 43% efficiency which is 7% higher than Q protocol but Q provides the preprocessing of next frames by processing one frame in one slot, whereas DBTSA does not proceed to next frame without finishing the process of current frame. So, it can be concluded that the DBTSA is idle for warehouse with high flow of RFID tags but when the slots are sitting for long and have less flow activity, Q protocol seems to be better than DBTSA. In terms of avoiding the rate of active tag collision, Q and DBTSA are more efficient to include tag in frames by reducing the collision rate.

484

A. Sukhwal et al.

7 Comparison of ALOHA Based Anti-collision Protocols As shown in Table 3. PA (Pure ALOHA), SA (Slot ALOHA), FSA (Frame Slot ALOHA), DFSA (Dynamic Frame Slot ALOHA) and Q algorithm are proven to be the best algorithmic protocols to favor tag collision avoidance in RFID system. The systematic comparison between these algorithm prove the DFSA as the most efficient algorithm to avoid the collision of the tag in any RFID system. On the other hand, DFSA possess the threat of having the size of frame more static, while the Q algorithm provides the dynamic frame allocation to make it more useful in case of handling high intense perimeter with the presence of huge number of tags and readers. By considering the complexity of Q and DFSA, the DFSA appears to be more complex than Q protocol. DFSA manipulates the frame size depending on the number of tags sending signals and reader changes function to change the frame size. This algorithm focuses on the reduction of slots by identifying and sorting the tags without crashing due to tag collision in such a way this algorithm focuses on the reduction of slots [1, 11].

8 Future of RFID (Radio Frequency Identification) Technology In 2014, the global market revenue of RFID technology was barely 9 billion US dollars. This has multiplied three times in a span of 8 years by making global market revenue to 27 billion US dollars as shown in Fig. 5. This strongly indicates the fact that RFID will continue to grow stronger in the industry and hence more and more research and development works will be initiated in future to improvise the problems existing within RFID systems such as tag and reader collision [16].

9 Conclusion Tag collision is one of the major concerns present in the field of RFID, Tag and Tag, Reader and Reader collision pose a threat to entire RFID system, which can be used for supply chain management or inventory tracking in warehouses. RFID also present opportunity in the global market due to its cost effectiveness and ability to integrate RFID technology into IoT. Along with that RFID provides different advantages such as high reliability, energy savings and low maintenance cost. This paper has successfully discussed about the RFID processes and proposes the most efficient and dependable tag anti-collision protocols. The extensive literature review carried out in this study demonstrates that a sufficient research and development has been conducted in this field. Whereas the current efficiency of the

RFID (Radio Frequency Identification) Tag Collision Risk Mitigation …

485

Fig. 5 RFID worldwide market revenue in billion U.S. dollars (credit statista) [18]

protocols indicates that more R&D is required to overcome the emerging RFID challenges. The key findings of the proposed study were that various algorithms and protocols have proven to be more efficient than the traditional methods over time, resulting in increased complexity. RFID must become more robust in the future, and anti-collision protocols play a critical role in this regard.

References 1. Chechi D, Kundu T, Kaur P (2012) The RFID technology and its applications: a review. Int J Electr Commun Instrum Eng Res Dev (IJECIERD). 2:109–120 2. Jung K, Lee S (2015) A systematic review of RFID applications and diffusion: key areas and public policy issues. J Open Innov 1:9. https://doi.org/10.1186/s40852-015-0010-z 3. Ibrahim AAA, Nisar K, Hzou YK, Welch I (2019) Review and analyzing RFID technology tags and applications. In: 2019 IEEE 13th international conference on application of information and communication technologies (AICT), pp 1–4. https://doi.org/10.1109/AICT47866.2019. 8981779 4. Assariana A, Khademzadeh A (2018) A beacon analysis-based RFID reader anti-collision protocol for dense reader environments. Comput Commun 128:18–34 5. Guo Z, Yuan J, Gu J, Liu Z (2010) Research of tag anti-collision technology in RFID system. In: Luo Q (eds) Advances in wireless networks and information systems. Lecture notes in electrical engineering, vol 72. Springer, Berlin. https://doi.org/10.1007/978-3-642-14350-2_29 6. Haoxiang W, Smys S (2021) A survey on digital fraud risk control management by automatic case management system. J Electr Eng Autom 3(1):1–14 7. Liu Z, Guan Z, Shang K, Cai W (2013) Anti-collision algorithm for RFID based on continuous collision detection. Telkomnika Indones J Electr Eng 11. https://doi.org/10.11591/telkomnika. v11i12.3661

486

A. Sukhwal et al.

8. Liu X, Yin J, Liu J, Zhang S, Xiao B (2022) Time efficient tag searching in large-scale RFID systems: a compact exclusive validation method. IEEE Trans Mob Comput 21(4):1476–1491 9. Yang Q, Liu X, Guo S (2021) No wait, no waste: a novel and efficient coordination algorithm for multiple readers in RFID systems. In: 2021 IEEE/ACM 29th international symposium on quality of service (IWQOS), pp 1–10 10. Pal K (2019) RFID tag collision problem in supply chain management. Int J Adv Perv Ubiquit Comput 11:1–12. https://doi.org/10.4018/IJAPUC.2019070101 11. Cmiljanic N, Landaluce H, Perallos A (2018) A comparison of RFID anti-collision protocols for tag identification. Appl Sci 8(8):1282. https://doi.org/10.3390/app8081282 12. Memon MQ et al (2018) Improving efficiency of passive RFID tag anti-collision protocol using dynamic frame adjustment and optimal splitting. Sensors (Basel, Switzerland), vol 18(4):1185. https://doi.org/10.3390/s18041185 13. Tan X, Wang H, Fu L, Wang J, Min H, Engels DW (2018) Collision detection and signal recovery for UHF RFID systems. IEEE Trans Autom Sci Eng 15(1):239–250. https://doi.org/ 10.1109/TASE.2016.2614134 14. Zhang J, Lyu Y, Patton J, Periaswamy SCG, Roppel T (2018) BFVP: a probabilistic UHF RFID tag localization algorithm using Bayesian filter and a variable power RFID model. IEEE Trans Ind Electron 65(10):8250–8259 15. Zhang D, Huang H, Jo M (2015) Future RFID technology and applications: visions and challenges. Telecommun Syst 58:193–194. https://doi.org/10.1007/s11235-014-9865-8 16. Nayak R (2019) Challenges and future directions of RFID technology. https://doi.org/10.1201/ 9781351238250-9 17. Sungheetha A (2021) COVID-19 risk minimization decision making strategy using data-driven model. J Inf Technol 3(01):57–66 18. Alsop T (2022) Global RFID technology market revenue 2014–2025. https://www.statista.com/ statistics/781338/global-rfid-technology-market-revenue/. Date of Publication: 2 Mar 2022

BizGuru 1.0: Design and Development of a Mobile-Based Digital Marketing Guide for Elderly Ahmad Sofian Shminan, Nur Zulaikha Mohamed Aziyen, Lee Jun Choi, and Merikan Aren

Abstract BizGuru 1.0 is an online learning platform using mobile devices known as mobile-based learning. It is a modernized alternative to acquiring knowledge which is suitable with the current digitalized environment. BizGuru provides learning materials that promote business-related knowledge, focusing on Digital Marketing. However, in this study, the mobile application design will be focusing on the elder’s group to cater for their needs. The target users are people aged 60 years old and above, who use an Android smartphone and are interested in gaining new knowledge. The purpose of the proposed application is to help these retired elderlies find an alternative that enables them to gain income at late age to continue supporting their living expenses. With the current pandemic situation and how they are often related to poverty, both circumstances result in the elders having to struggle to survive financially. Therefore, by using BizGuru, the elderlies do not only get to familiarize themselves with modern devices, but also they could look for other alternatives to gain income and avoid poverty which helps to fulfil the 1st goal of Sustainable Development Goals (SDG) on the eradication of poverty issues. Besides, this proposed application also provides learning opportunities for elderlies who have the desire to gain knowledge at late age which can help fulfil the 4th goal of SDG which is promoting life-long learning opportunities for all. Keywords Mobile-based learning · Digital marketing · Elderly community

A. S. Shminan (B) · N. Z. M. Aziyen · L. J. Choi · M. Aren Faculty of Cognitive Sciences and Human Development, Universiti Malaysia Sarawak, 94300 Kota Samarahan, Sarawak, Malaysia e-mail: [email protected] N. Z. M. Aziyen e-mail: [email protected] L. J. Choi e-mail: [email protected] M. Aren e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_37

487

488

A. S. Shminan et al.

1 Introduction The world is now becoming more digitalized with the aid of Information and Communication Technology (ICT) in the forms of mobile phones, computers, laptops, tablets, and TVs until it becomes a necessity in everyday life. ICT has upgraded human’s life such as transforming handwritten letters to electronic mails, in-store shopping to online shopping, and traditional classroom to online learning [1]. This fast-paced technology advancement took place due to the impact of the Covid-19 pandemic which has also increased the e-commerce trends worldwide, including Malaysia. The extensive growth of e-commerce is because of the fact that people are mostly adapting to the new norm which is keeping a safe social distancing and staying indoors to stop the spreading of the virus [2, 3]. The gradual development of ICT in human life since years ago also indicates the significance of technologies in improving the quality of life. With the aid of devices, the productivity and performance of an organization are able to grow positively. However, despite the economic growth in developing countries like Malaysia, people who are much older, or the elderly group, are found to be at risk of poverty in later life [4]. An improper retirement planning is one of the factors that contribute to the rising number of poverty issues among retired elderlies. Due to the increasing life expectancy among the elderly, the retirement planning is very crucial to support their living expenses after retirement [5]. In fact, their only source of income after retirement will be the money that they earn from the past few years of working and the monthly allowance from their family members who are still working. Therefore, to overcome the issues and help the elderlies to reduce their financial burden, a mobile-based learning platform called ‘BizGuru’ is proposed. Setting up a small business is a great idea to gain income at late age. With the knowledge and guidance on digital marketing provided in BizGuru, the elderlies could promote their businesses on various online platforms. A digital marketing is used to allow the potential customers to learn about the product or service offered. According to Sharma [6], digital marketing is proven to be more effective than traditional marketing due to the growing number of people using digital technologies to shop. Therefore, companies that do not include digital marketing strategies in their businesses such as using blog to market their products are at loss [7]. To fulfill the 4th goal of Sustainable Development Goals (SDG) on promoting lifelong learning opportunities [8], the development of mobile-based learning platform is to encourage the elderlies to gain knowledge despite their old age. This learning platform also helps to create the sense of belonging among the elderlies in the digitalized society. This mobile-based learning platform consists of learning materials on digital marketing in the form of texts, visuals, and audios. With this proposed application, it helps to provide the people interested in digital marketing, the skills and knowledge which they could implement in businesses. Somehow, it will encourage the elderlies to open a small business to help them gain side income, and further to eliminate poverty issues at old age and next, to fulfill the 1st goal of SDG on eradicating poverty. This article is organized in such a way that the following section

BizGuru 1.0: Design and Development of a Mobile-Based Digital …

489

gives some related work, whereas the outline of our proposed BizGuru is given in Sect. 3 thereon. The Conceptual Process Flow and Model of BizGuru are described in Sect. 4, respectively. Section 5 gives explanations on findings and discussions. Finally, the conclusions are drawn and remarks are made in the last section.

2 Related Work The development of a mobile-based learning application for elderlies are quite few in numbers. This is due to the needs to be catered for elderlies in terms of UX design are a bit complex as they have poor health conditions. However, these are several past studies that can be found relatable with the proposed application.

2.1 Smartphone for Seniors (S4S) Project In year 2013, Barros et al. [9] had developed a health-related mobile-based learning application for older adults where they can learn to work out for fall prevention. They conducted the usability test at a local adult care centre for 3 sessions, as an improvement on the app was needed after each session to get the best results. This study shows that visual design is one of the most crucial elements in designing a mobile learning application for older adults. People who have their way of perceiving things, including older adults that are not familiar with the use of a mobile application, have a higher tendency to perceive the functional buttons wrongly. To overcome this problem, the use of icons to represent the text in a button is recommended in this study. It will help to improve the cognitive affordance of an element in a mobile application and avoid errors. Therefore, buttons with short text descriptions and representative icons will be included in the new mobile based learning application to improve the older adults’ experience with technology.

2.2 Enhancing Islamic Knowledge via Short Messaging System (SMS) Alkasirah and Nor [10] had conducted experimental research to explore the potential of mobile-based learning using SMS, based on Adult Learning Theory (Andragogy). The learning topic is about ‘waqafa’ or known as ‘waqf’, an Islamic knowledge. The adult learners were tested using the Solomon Four Group Design method and ‘waqf’ questionnaire whereby the learners were divided into 4 groups (2 groups are control groups, and another 2 groups are the experimental groups). Only the control groups received the SMS regarding ‘waqf’ knowledge. A pretest and posttest were

490

A. S. Shminan et al.

conducted and the results have shown that the experimental group manages to score higher marks in the posttest which supports the use of mobile learning in education. Besides that, another finding in this research is that adult learners are self-directed as mentioned in the Adult Learning Theory (Andragogy). Hence, to develop a mobilebased learning application for adult learners, developers need to design content with easy comprehension such as avoiding the use of high vocabulary words. Using high vocabulary words could break the adult learners’ self-esteem and lead to the lack of use of the mobile-based learning application.

2.3 eVideo Mobile Patzer et al. [11] had researched a game-based work- related e-learning course which is a digital media in the hospitality industry known as, eVideo. The eVideo used to be web-based learning only. However, with the advancement of technology nowadays that supports mobile-based learning, eVideo is redesigned to be supported by smartphones. The eVideo mobile consists of learning materials including audio and subtitles which are developed by referring to Universal Design for Learning (UDL) and exercises with multiple-choice questions. A pilot study was conducted, and the results have shown that the adult learners were able to adapt to the transformation from web- based to mobile-based learning. Besides, eVideo mobile also made the learning contents accessible anytime and anywhere. However, the downside of the mobile-based is due to the size of the screen compared to web-based. Therefore, to overcome this problem, the new mobile-based learning should be designed using texts with an appropriate size and suitable font for adult learners.

2.4 The SenApp Project Leen-Thomele et al. [12] had developed SenApp to evaluate the mobile learning concept on older adults in France and Germany. The SenApp focuses on educating older adults on how to use the Skype application and e-mail. There is a total of 12 short learning materials that explain the communication application. A pilot study was conducted and a total of 40 participants (20 learners from each country) was involved. The older adult learners were required to fill out the questionnaires after finishing the learning materials module. The learning materials were rated good by the learners although they still require some improvement. Older adults’ learner with low education background gives a low rating on the SenApp because they have a lack of experience in e-learning and the learning materials are quite hard to be understood. This leads them to spending a shorter time on learning due to low self-confidence and motivation. Therefore, to build a new mobile-based learning application for older

BizGuru 1.0: Design and Development of a Mobile-Based Digital …

491

adults, the variables such as different target users’ educational background should be considered. By reducing the amount of learning material for mobile-based learning, the older adult learners could have a better learning experience with less stress and help them to cope with technology advancement.

2.5 iPractice: Tablet-Based Home Practice Program in Aphasia Treatment Kurland et al. [13] had conducted a pilot test on the effectiveness of a tablet-based home practice program in aphasia treatment. The application used is known as ‘iPractice’ which is specifically designed for iPad users. The participants recruited in this study are 55–81 years of age, with language deficiency due to past accidents that had affected the brain responsible for linguistics. Since the participants are at old age, the mobile application is designed with simplicity to provide ease of use in navigating while using the application for people not familiar with the technology. The ‘iPractice’ application contains 2 types of interactive books: one with objects and one with actions, and each book contains 20 words for the participants to practice. Half of the number of words (10 words) were trained in the treatment program and the others were not trained. Due to the participants’ deteriorating eyesight due to old age, the videos and tasks given were able to be viewed in full screen. The participants were able to view the video unlimited times during self-practice. The minimum required time to practice was 20 min in 5 or 6 days every week. After every 6 months, the participants were required to follow a check-up at the clinic to check the progress at least once. The results of this study were positive, as they managed to improve their linguistic ability, although there were some complaints received from the users due to their boredom in using the application for practicing for a long period of time. Therefore, in the proposed application designed for the elderlies, each learning material should be designed with simplicity to reduce the time taken for learning and to avoid the feeling of irritation, which will lead to the application being unused.

2.6 Reflection Therefore, by referring from the previous studies relating to mobile-based learning on the elderlies [7–11], it shows that the elderlies are able to gain knowledge despite their old age. In fact, with the mobility of mobile-based learning, it gives the opportunities for learners to learn at any time and any place. However, there are disadvantages of mobile-based learning on this group such as poor eye-sight and lack of familiarity with the devices. Therefore, to overcome this issue, developers need to implement

492

A. S. Shminan et al.

the UX Design knowledge [14] and align with self-directed learning theory [15] to cater for the needs for learners at old age and provide sufficient information for them to navigate with ease while using the proposed mobile application.

3 Methodology Instructional System Design (ISD) is a structured development of a digitalized learning platform that applies the instructional theory to establish a set of effective instructions to aid the learning process. In other words, it functions to help people gain knowledge successfully using technology devices [16, 17]. A well-designed tool for an educational purpose helps to improve the learner’s ability to learn and to become self-directed [18]. Therefore, developers used ISD models as a reference to plan, create high-quality instructions, and build the system. Besides, a well-known ISD model that is commonly used in designing and developing system is ADDIE model [19]. ADDIE model consists of 5 phases: analysis, design, development, implementation, and evaluation. Each of these phases in Fig. 1 gives out a necessary result to move to the following stage of the development process [20, 21].

3.1 Analysis Target Audience Target audience is the group of people who are identified to likely be interested in the product offered. BizGuru application design is focused on catering to the needs of old learners. Therefore, the target audience in this study is a group of retired elderlies aged 60 years old and above, regardless of gender and ethnicity. The criteria are that they must be interested to learn new knowledge and own an Android smartphone. A stable internet connection is required to download and install BizGuru. Apart from that, this target group should have a big interest in learning digital marketing skills or improving their business strategy. Gap Analysis The elderlies are found to be at risk of poverty at later life [2, 3]. Therefore, to help the elderlies find an alternative to gain income at old age, a mobile-based learning application on digital marketing is proposed. This will allow the elderly to adapt with the current technology and help them earn money using an online platform which is easier and which needs less energy and physical movement. Desired Outcome By exposing the elderlies to digital marketing, they may be encouraged to run a small business to support their living expenses. The elderlies can promote their products or services made from the skills or creativity that they had learned personally throughout

BizGuru 1.0: Design and Development of a Mobile-Based Digital …

493

Fig. 1 ADDIE model flowchart

their long lives. As a result, the number of aged people with financial problems or at risk of poverty can be reduced and this, in the long run, can provide a better quality of ageing life [8]. Besides that, the elderlies would also get to prevent old-age disease by learning new knowledge during their leisure time, as a form of brain exercise.

494

A. S. Shminan et al.

Fig. 2 Learning screen

3.2 Design The high-fidelity prototype of BizGuru was made using Adobe XD software which is well-known for its impressive user interface (UI) and user experience (UX) designing and tools to build the prototype for mobile applications that allow a smooth process of moving from static mockups to interactive prototypes [22, 23]. The Bahasa Melayu version has a similar interface design with English version except for the language usage in audio, video, and writing. Apart from that, the course purpose, course flow, learning materials, and strategy to deliver the knowledge is figured in this stage. The course flow and learning materials were referred from a well-known book in the digital marketing industry titled ‘Digital Marketing for Dummies’ by Deiss and Henneberry [24]. The strategy to deliver knowledge successfully is by incorporating visual presentation and audios are shown in Fig. 2. Since this proposed mobilebased learning application is meant for elderlies whereby it is common for them to have poor eye sight, the design interface of each screen will include the UX Design principle to ensure that they get the best learning experience [14].

3.3 Development Ionic5 was used as the programing tool to build the real application. Ionic5 is known for its cost-effective development with high speed performance and functionality, it has an open source feature, and uses simple programming language. This software uses Hyper Text Markup Language (HTML), JavaScript (JS) and Cascading Style Sheet (CSS). Next, Visual Studio Code (VSC) software was chosen as the platform

BizGuru 1.0: Design and Development of a Mobile-Based Digital …

495

to write and compile all the coding built for BizGuru interface design and functionality. The pages created in VSC consist of a subpage focusing on the layout and placement of element using HTML, another subpage focusing on the design, decorations, and styling using CSS, and subpages focusing on the functionalities using JS. Also, Windows Powershell was also used in giving commands for development purpose such as to include the Android Studio extension in Ionic5 project for testing the developed application using android emulator and build APK file of BizGuru. This APK file is a type of an executable file which is used to compile the mobile application and allow both the distribution and installation of application. Lastly, BizGuru learning materials as shown in Fig. 3, such as infographics and high-quality animation videos for every topic were developed using Canva and Animaker online software.

Fig. 3 Application setup with learning assessment

496

A. S. Shminan et al.

3.4 Implementation As the cases of Covid-19 pandemic is steadily rising, the data collection procedure had to be transformed into the online method. Therefore, two different websites were created using cloud website builder that is meant for expert evaluation and pilot and usability test. These websites helped to organize the documents and materials needed to conduct the evaluation session. The respective website link was given to the evaluators and participants through e-mail or WhatsApp message. In the website, there is an introduction page for explaining the concept and purpose of study, briefing notes and instruction to perform the required task, informed consent form to explain the terms and conditions after they had agreed to participate in this study, a video demonstration of BizGuru app, a Google Drive link containing an APK file of BizGuru along with the poster guide to install the application, and a link to two different online evaluation forms created using Google Form to collect the ratings and comments given by the evaluators and participants.

3.5 Evaluation There are three phases of evaluation session done in this study namely expert evaluation, pilot testing, and usability testing. Expert evaluation is a type of assessment that relies on the individual expertise in that particular area of study to discover the potential issues that have arisen in the mobile application [25]. Meanwhile, the pilot testing is an additional step that acts like a rehearsal phase with a smaller number of participants before conducting the main study (such as usability testing) with a larger number of participants [26, 27]. Usability testing is the main research study aiming to test the system developed in real environment which requires the participants to perform, complete a task given, and provide feedback to the researcher [27]. BizGuru application overall performance was evaluated from an expert’s and user’s point of view. 3 expert evaluators with a strong foundation and knowledge in the technicality and usability of a mobile application interface design were recruited in this study. The instrument used in this evaluation session is known as Mobile Learning Usability Attribute Test (MLUAT) [28, 29]. Other than that, 7 retired elderlies (2 elderlies in pilot test, 5 elderlies in usability testing) had volunteered to participate in this study where they share a common characteristic of being an Android smartphone user. However, these elderlies are not experts in this field of study, as opposed to the expert evaluators. In fact, these elderlies are the target users who will use, learn and gain the knowledge from the mobile learning application developed. The instrument used in this evaluation session is known as System Usability Scale (SUS). The online evaluation form for MLUAT and SUS was created using Google Form with a 5-likert scale answer. The benefit of using 5-likert scale instead of 7 or more Likert scale is because it helps to produce a reliable quantitative data that is easy to be analysed, a universal method of data collection, and ideal method for a long

BizGuru 1.0: Design and Development of a Mobile-Based Digital …

497

questionnaire with multiple choice. Their ability to understand the functions in the mobile-based learning application was included in the feedback form. The usability and ease of use of the mobile-based learning application was measured throughout the entire interaction process and the feedback on the overall effectiveness of the proposed application to deliver the knowledge given.

4 Conceptual Process Flow and Model of BizGuru A conceptual model is used to represent the whole process of a system. Flowchart is one of the best ways to show the overall cycle with a clear sequence of actions. Figure 4 shows the flowchart of BizGuru mobile-based learning application. The BizGuru mobile-based learning application starts off with a splash screen welcoming the learners to use this application. Next, the learners need to choose their preferred language for learning. There are two different languages included in this application which is Bahasa Melayu and English. After choosing the language, the learners will be directed to another screen which shows all the learning topics available in this application ranging from Topic 1 to Topic 5. To start the lesson, the learner can choose any of the learning topics, preferably starting from the first topic as it has a hierarchical order. On each learning topic, the learner will be provided with a short overview to give a brief idea of the learning topic and a video that will help in providing a more detailed explanation. After watching the video, an infographic is provided on each topic which will help the learners to summarize their understanding in visual and textual forms. After the learning process, the learner will be assessed with a short quiz to evaluate their level of understanding. The quizzes will be given after each learning process with a ‘Yes’ or ‘No’ answer. Then, the learners will obtain their results on the spot with the feedback that was set up to be provided after the learner tapped on the answer button. If the learner wishes to improve their understanding, they are allowed to retake the learning topic or otherwise, and then proceed to the next learning topic. In the next learning topic, the learning flow is all the same. Lastly, learners are allowed to exit or access the learning materials at any time. Figure 5 shows the conceptual design of BizGuru mobile-based learning application. The target users of the proposed application are elderlies who seek for a lifelong learning opportunity. Firstly, in the training phase, the learners will access the learning materials by reading a brief overview of the learning topic. Then, as the elderly face some difficulties reading on the small screen for a long period of time, a video explaining on the related topic will be provided after each topic overview. The group will then use their senses to capture the information shown on screen. The information captured through the hearing sense will be kept in the working memory on verbal mode. Meanwhile, the information captured using the eyes will be kept in the working memory on pictorial mode. The valuable information will be interpreted into knowledge and kept in the long-term memory. Next, the learners will undergo

498

Fig. 4 Conceptual flow process

A. S. Shminan et al.

BizGuru 1.0: Design and Development of a Mobile-Based Digital …

499

Fig. 5 Conceptual model

an assessment phase which requires them to answer a short quiz to test their understanding and evaluate the effectiveness of the mobile- based learning application in delivering knowledge. These quizzes are provided at the end of each learning topic. The answers chosen will be verified and the learners will obtain their results immediately in the form of pop-up feedback. Lastly, after completing the whole course, the learners can use the knowledge learned to apply to real cases. Learners can start their small online businesses using Instagram and promote their products using the digital marketing knowledge.

5 Findings and Discussion Evaluation process proves to be a significant step in creating a new application system. The main notion of the system testing is to ensure that the system launched can fulfil the user requirement and satisfaction successfully. The testing must be conducted prior to the proposed application development to ensure that a better product can be obtained before the mass production.

5.1 Subject Matter Expert The subject matter expert (SME) offers sound knowledge and expertise in a specific subject, business area, or technical area for a project [25, 30]. SME also makes sure that the facts and details are correct so that the project’s/program’s deliverable(s) will fulfil the stakeholders’ needs, also the requirements of the legislation, policies,

500

A. S. Shminan et al.

standards, and best practices. To assess the application, expert received the online link to do the testing on the application. The expert then answered the questionnaire provided. From the finding, the expert evaluator was given the questionnaire to answer the MLUAT evaluation. There are six parts in the evaluation namely visibility status, leaner control and freedom, match between the system and real world, consistency, error prevention, recognition rather than recall and minimize information on screen. The total score from all three expert evaluator feedback was calculated to be 95.93 and the average total score for BizGuru was 4.36. By referring to the 5-likert scale used, low values such as 1 and 2 represent a negative attitude meanwhile higher numbers such as 4 and 5 represent a positive attitude [31]. The average score (4.36) belongs to the higher value group which confirms that BizGuru mobile learning application had acquired a positive attitude from the expert evaluators in the expert evaluation phase, implying a good functionality and usability of the application.

5.2 Usability Evaluation Usability testing and System Usability Scale (SUS) are the technique and instrument used in the user testing phase- the practice of testing is the extent of the ease of use of the design, to be applied to a group of representative users. It entails an observation of users as they try to complete tasks and receive feedback from the users by way of interviews or questionnaires about user satisfaction on the product’s prototype. According to Macefield [27], 5–10 respondents are the least number of respondents required for the usability testing. Nielson stated that elaborate usability tests denote a waste of resources, as the best results come from testing not more than 5 users and running as many small tests as possible [32]. This is due to the fact that when more tests are run on the user, the same results are to be obtained as the previous users where the first study with 5 participants is considered sufficient to find 85% of the usability problems. Therefore, 5 participants from elderly community are chosen at random to be part of the usability testing [27]. The feedback obtained will be gathered and analysed to be further improved in the application development. The SUS score per user was calculated to be 72.5, 77.5, 70, 72.5, and 75. All five users gave a rating of Agree [4] in question number 1 where they admitted that they would use BizGuru frequently. On the other hand, they also gave a significantly low rating on items depicted in questions 8 and 10 where they disagreed to the statement that BizGuru application was cumbersome to use, and they also disagreed to the statement whereby they needed to learn a lot of things before using BizGuru application. These questions are illustrated as negative questions, thus they received low ratings, depicting that BizGuru is easy to use. Besides, the average SUS score for BizGuru was calculated to be 73.5 (Grade B) which is considered as a good application, although it still requires an improvement on certain areas.

BizGuru 1.0: Design and Development of a Mobile-Based Digital …

501

5.3 Related Theory and Methodological The Self-directed learning and Andragogy theory was embedded in developing this application. Both theory defines adult learning as unique and different from general learning. It remarks that the elderly was classed as independent learners since they have a sense of responsibility from their life experiences. By looking at the facts, adults spend more time acquiring specific skills, which can be performed through reading, listening, observation, reflection, and exercise [15]. BizGuru 1.0 is an efficient device that combines learning material related to quail farming in a single platform. It is believed to be more efficient than the traditional website or book as the user can readily download the application and learn independently, which is referred from the theories embedded. The SDL theory describes a process where an individual takes the initiative in learning, including the learning materials and strategies, and evaluates the outcomes. It is opposite to teacher-directed learning or “pedagogy,” which is considered an information transfer influenced by the outer in determining the learning outcome. According to Farage et al. [14], the age factor plays a role in the physical changes of a person; therefore, a design guideline is referred to achieve the satisfaction level of users. In this study, to meet the elderly needs, the universal design principle related to the elderly is referred to in measuring the achievement of the development application. The most common decline in the elderly is the visual field; thus, color choices are essential in accommodating age-related visual impairment. The longwavelength or “warm” colors are preferred to convey information that differs from the background. Simplicity is the word best defining the visual display, where the essential information is presented in a noticeable and uncrowded field. Short and precise learning materials matter as the degree of working memory in the elderly is shorter than in young adults. An application with complex information will create an overload of information to be processed for the elderly. ADDIE, an Instructional System Design method [19] was embedded in the development of BizGuru 1.0. It is intended to solve a problem by developing and evaluating IT artefacts ranging from software, formal logic, or natural language. In this study, the issue of poverty among the elderly is chosen. ADDIE involves a process that enables an understanding of the problem addressed and the feasibility of the possible solution. BizGuru was made to disseminate information on digital marketing efficiently and to ease the learning for the target user (the intention and usefulness). In the hope that this application would help the target user to gain income from initiating their quail farming and business (the benefits). It aligned with the ADDIE method concerning the artifacts’ intention, usefulness, and benefits. The expertise and target user were involved implicitly during the evaluation to receive direct feedback and to test the perception of the usefulness of proposed mobile based learning application. The most time-consuming process in this study is design and development, which is carried out non-stop even after evaluation. The feedback received during the evaluation will be considered, and changes to the application will be made.

502

A. S. Shminan et al.

6 Conclusion and Future Work BizGuru is developed from the analysis of the weaknesses or disadvantages found from previous mobile learning application studies conducted from 2014 to 2019. The improvement has been made to provide a better learning experience for the old learners which also help to boost their cognitive skills and well-being by experiencing what is termed as ‘quality ageing’. Besides that, the development of BizGuru which focuses on tending to the needs of the aging learners also helps to fulfil the 4th goal of Sustainable Development Goals (SDG) that is to provide a life-long learning opportunity [8]. As this mobile learning application promotes a digital marketing knowledge, the retired learners can take this opportunity to make it as an alternative to gain side income to continue supporting their living expenses in later life using technology devices. This benefit also helps to fulfil the 1st goal of SDG that is to eradicate poverty issues among the elderly group. Besides, there are a few limitations found in this study- for one, conducting the usability testing remotely amid the Covid19 pandemic had caused the application designer to be unable to observe the users directly during the testing and evaluation process to get a more detailed remark and experience. There was also a language barrier issue, as one of the users in the usability testing phase had commented that it would be better if there is a Chinese version in the BizGuru mobile learning application. This is due to the differences of races available in Malaysia and the daily language usage by the participants. Hence, future studies are recommended to improve the functionality of the mobile learning application by connecting to a database which will allow a bigger space for learning contents and also record the learners’ progress. By implementing a database architecture [33], it will also allow the application developers to monitor the learners’ learning curves based on their achievements in the quiz section. Apart from that, the future mobile learning application should improve the cloud security features by adding specific mechanism for a slot registration and log-in page [34]. This will instill the learners with a sense of belonging and security, as their learning progress is kept confidential to themselves without exposure to other learners. Not all learners are comfortable sharing their achievements with other learners, especially the competitive and introvert ones. Lastly, by having an authentication page, it allows multiple users per device, and this means that the number of learners and users of the respective mobile learning application can increase.

References 1. Rothe F-F (2020) Rethinking positive and negative impacts of ‘ICT for development’ through the holistic lens of the sustainable development goals. Inf Technol Dev 26:653–669. https:// doi.org/10.1080/02681102.2020.1756728 2. Galhotra B, Dewan A (2020) Impact of COVID-19 on digital platforms and change in Ecommerce shopping trends. In: 2020 fourth international conference on I-SMAC (IoT in social, mobile, analytics and cloud) (I-SMAC), pp 861–866

BizGuru 1.0: Design and Development of a Mobile-Based Digital …

503

3. Winarsih, Indriastuti M, Fuad K (2021) Impact of Covid-19 on digital transformation and sustainability in small and medium enterprises (SMEs): a conceptual framework. In: Barolli L, Poniszewska-Maranda A, Enokido T (eds) Complex, intelligent and software intensive systems. Springer International Publishing, Cham, pp 471–476 4. Masud J, Hamid T, Haron SA (2015) Measuring poverty among elderly Malaysians 5. Mohidin R, Jamal AAA, Arokiadasan CG et al (2013) Revisiting the relationship between attitudes and retirement planning behavior: a study on personal financial planning. Int J Multi Thought 3:449–461 6. Sharma DU, Thakur PKS (2020) A study on digital marketing and its impact on consumers purchase. Int J Adv Sci Technol 29:13096–13110 7. Bala M, Verma D (2018) A critical review of digital marketing. Social Science Research Network, Rochester 8. Poverty Eradication | Poverty Eradication. https://www.un.org/development/desa/socialperspe ctiveondevelopment/issues/poverty-eradication.html. Accessed 6 Jun 2022 9. de Barros AC, Leitão R, Ribeiro J (2014) Design and evaluation of a mobile user interface for older adults: navigation, interaction and visual design recommendations. Proc Comput Sci 27:369–378. https://doi.org/10.1016/j.procs.2014.02.041 10. Alkasirah NAM, Nor NMM (2019) Potential usage of mobile learning via short messaging system (SMS) for enhancing Islamic knowledge of adult learners 11. Patzer Y, Lambertz J, Schulz B, Pinkwart N (2018) Mobile online courses for the illiterate: the eVideo approach. In: Miesenberger K, Kouroupetroglou G (eds) Computers helping people with special needs. Springer International Publishing, Cham, pp 379–383 12. Leen-Thomele E, Hetzner S, Held P (2016) Mobile learning concepts for older adults: results of a pilot study with tablet computers in France and Germany. In: Zhou J, Salvendy G (eds) Human aspects of IT for the aged population. healthy and active aging. Springer International Publishing, Cham, pp 319–329 13. Kurland J, Wilkins AR, Stokes P (2014) iPractice: Piloting the effectiveness of a tablet-based home practice program in aphasia treatment. Semin Speech Lang 35:51–64. https://doi.org/10. 1055/s-0033-1362991 14. Farage MA, Miller KW, Ajayi F, Hutchins D (2012) Design principles to accommodate older adults. Glob J Health Sci 4:2–25. https://doi.org/10.5539/gjhs.v4n2p2 15. Roberson DN (2004) The nature of self-directed learning in older rural adults. Ageing Int 29:199–218. https://doi.org/10.1007/s12126-004-1017-0 16. Shin S, Bae Y (2015) A study on the hierarchical instructional system design of software education by school system. J Korean Assoc Inf Educ 19:533–544. https://doi.org/10.14352/ jkaie.2015.19.4.533 17. Shminan AS, Choi LJ, Sharif S (2020) AutiTEACCH: Mobile-based learning in a structured teaching approach for autistic children caregivers. In: 2020 international conference on informatics, multimedia, cyber and information system (ICIMCIS), pp 259–264 18. Uz R, Uzun A (2018) The influence of blended learning environment on self-regulated and self-directed learning skills of learners. Eur J Educ Res 7:877–886 19. Valverde-Berrocoso J, Fernández-Sánchez MR (2020) Instructional design in blended learning: theoretical foundations and guidelines for practice. In: Martín-García AV (ed) Blended learning: convergence between technology and pedagogy. Springer International Publishing, Cham, pp 113–140 20. Magliaro SG, Shambaugh N (2006) Student models of instructional design. Educ Tech Res Dev 54:83–106. https://doi.org/10.1007/s11423-006-6498-y 21. Castro MDB, Tumibay GM (2021) A literature review: efficacy of online learning courses for higher education institution using meta-analysis. Educ Inf Technol 26:1367–1385. https://doi. org/10.1007/s10639-019-10027-z 22. Figliolia AC, Sandnes FE, Medola FO (2020) Experiences using three app prototyping tools with different levels of fidelity from a product design student’s perspective. In: Huang T-C, Wu T-T, Barroso J et al (eds) Innovative technologies and learning. Springer International Publishing, Cham, pp 557–566

504

A. S. Shminan et al.

23. Stoeva M (2021) Model and prototype of interactive assistant for compliant interface development—MayUI tool. In: 2021 international conference automatics and informatics (ICAI), pp 295–300 24. Deiss R, Henneberry R (2017) Digital marketing Fd, 1st edn. For Dummies, Hoboken 25. Perez RS, Fleming Johnson J, Emery CD (1995) Instructional design expertise: a cognitive model of design. Instr Sci 23:321–349. https://doi.org/10.1007/BF00896877 26. Shminan AS, Choi LJ, Barawi MH et al (2021) InVesa 1.0: the conceptual framework of interactive virtual academic advisor system based on psychological profiles. In: 2021 13th international conference on information and communication technology and system (ICTS), pp 112–117 27. Macefield R (2009) How to specify the participant group size for usability studies: a practitioner’s guide. J Usability Stud 5:34–45 28. Fetaji M, Fetaji B (2011) Comparing developed MLUAT (mobile learning usability attribute testing) methodology with qualitative user testing method and heuristics evaluation. In: Proceedings of the 12th international conference on computer systems and technologies. Association for Computing Machinery, New York, pp 516–523 29. Arain AA, Hussain Z, Rizvi WH, Vighio MS (2016) Evaluating usability of M-learning application in the context of higher education institute. In: Zaphiris P, Ioannou A (eds) Learning and collaboration technologies. Springer International Publishing, Cham, pp 259–268 30. Le Maistre C (1998) What is an expert instructional designer? Evidence of expert performance during formative evaluation. ETR&D 46:21–36. https://doi.org/10.1007/BF02299759 31. Norman G (2010) Likert scales, levels of measurement and the “laws” of statistics. Adv in Health Sci Educ 15:625–632. https://doi.org/10.1007/s10459-010-9222-y 32. Bangor A, Kortum PT, Miller JT (2009) Determining what individual SUS scores mean: adding an adjective rating scale. Undefined 33. Mishra A, Jaiswal A, Chaudhari L, Bodade V (2022) Health record management system—a web-based application. J IoT Soc Mob Anal Cloud 3:301–313 34. Shakya S (2020) Survey on cloud based robotics architecture, challenges and applications. JUCCT 2:10–18. https://doi.org/10.36548/jucct.2020.1.002

Development of Secure Cloud-Based Healthcare Management Using Optimized Elliptic Galois Cryptography V. Gokula Krishnan, D. Siva, S. MuthuSelvi, T. A. Mohana Prakash, P. A. Abdul Saleem, and S. Mary Rexcy Asha

Abstract The ever-increasing amount of e-medical data poses a security risk because of technological advancements in the healthcare business. An unstructured and large amount of unstructured data is generated by the healthcare data management system because of the wide variety of data formats that are used to capture patient information. Branches of hospitals can also be found in different parts of a city or state. Health information on patients that is kept in multiple places must be merged from time to time for research purposes. Cloud-based healthcare management systems can be an effective solution for storing and managing health care data more V. Gokula Krishnan (B) Department of CSE, Saveetha School of Engineering, Saveetha Institute of Medical and Technical Sciences (SIMATS), Thandalam, Chennai, Tamil Nadu 602105, India e-mail: [email protected] D. Siva Department of Computer Science, Faculty of Science and Humanities, SRM Institute of Science and Technology, Ramapuram, Chennai, Tamil Nadu 600089, India e-mail: [email protected] S. MuthuSelvi Department of CSE, Vel Tech Multi Tech Dr Rangarajan Dr Sakunthala Engineering College, Avadi, Chennai, Tamil Nadu 600062, India e-mail: [email protected] T. A. Mohana Prakash Department of CSE, Panimalar Engineering College, Poonamallee, Chennai, Tamil Nadu 600123, India e-mail: [email protected] P. A. Abdul Saleem Department of CSE, CVR College of Engineering, Mangalpally, Hyderabad, Telangana 501510, India e-mail: [email protected] S. Mary Rexcy Asha Department of IT, Panimalar Engineering College, Poonamallee, Chennai, Tamil Nadu 600123, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_38

505

506

V. Gokula Krishnan et al.

effectively. However, security is the most pressing issue with a cloud-based healthcare system. Elliptic Galois Cryptography (EGC) is used in this study to encrypt medical data files, and the value of the Galois field is determined using the Mayfly Algorithm. As a result, the proposed model is referred to as a “optimal EGC”. Use of the elliptic curve over a Galois field in elliptic curve cryptography reduces rounding errors. The healthcare data is protected in terms of both confidentiality and integrity when it is shared via the health cloud. Experiments have shown that the ideal solution can be computed more quickly in terms of file upload and download speeds as well as key generation and generation time. Additionally, it protects healthcare data from being tampered with during transmission via the health cloud. Keywords Cloud computing · Security · Key generation · Mayfly algorithm · Elliptic Galois cryptography · Medical data

1 Introduction When it comes to health care, the wearable technology is kindly spreading its tentacles to include not only all walks of hominid life, but also challenge the stowed huge health care data [1]. There are many different kinds of health care data, and they are made at a high rate, making it challenging to keep them locally. So the need for medical media applications like multimedia email, presentations, high quality audio and video sharing and shared papers has increased exponentially [2, 3]. In the healthcare industry, all patient records must be stored in the Cloud for future reference. This study examines the day-to-day operations of the health care business. Computational and processing challenges are plaguing the current health care business. Physical storage, security and medical errors are inherent issues in the traditional healthcare industry. It is critical to keep patient records safe since they include sensitive information. Patient data is being compromised by a number of issues in the current system. It takes up a lot of memory space, which is not cost-effective [4, 5]. In order to protect patient information, the cloud offers a high level of security. Prescription retrieval is easy because the patient’s info is stored in the cloud, so they may access it whenever they want [6]. Because the data is stowed in the cloud, anyone with a mobile device, such as a smartphone or PDA, can access it without requesting specific permission. One of the most dynamic sectors of the information technology business is the healthcare sector, where cloud computing is becoming increasingly important [7, 8]. Internet-enabled devices can access health-care information throughout the world thanks to cloud computing technology. The medical community can also benefit from the exchange of resources and information with other leading researchers in the same subject around the globe. To improve and develop the current health care industry, this study is being conducted. Anuradha et al. [9] Despite the advantages of cloud-based health care schemes, many doctors and healthcare institutions are reluctant to utilize them because of the risk of data

Development of Secure Cloud-Based Healthcare Management Using …

507

breach. Also, because of the subtle nature of the data being kept and retrieved, numerous health care organizations are avoiding public cloud and installing private cloud services in its place [10]. The electronic health annals must be securely transferred via networks in order to protect patient confidentiality and data integrity. In [11, 12], the drawbacks of using a cloud database to store health information are discussed. For the protection of electronic health records during transmission, a cryptosystem is typically needed. Security of user data has been achieved by the employment of conventional methods such as the Rivest–Shamir–Adleman (RSA)-based system [13, 14], and the encryption of user data [15]. Large keys and complicated computations make it difficult to use such systems on mobile devices, which is another drawback of these approaches. The use of Elliptic Curve Cryptography (ECC) in cryptosystems has grown in recognition and use during the past several years. In order to achieve great scalability and efficiency, ECC reduces the key complexity by using smaller key lengths. In order to store and share health data in the cloud, users must encrypt their data before uploading it to the servers. The health cloud will benefit from this research because of the optimized EGC implementation, which will allow it to provide better healthcare services while maintaining the integrity of patient data. In order to build an Access Control List (ACL), the TTP-CS receives healthcare data, a list of CUs, and the necessary criteria from the data owner. Later, the encrypted data is transferred to the HC on behalf of the CU for storage. If the CU is interested in accessing health data files, the TTP-CS will get a download request. The following are a few of the methodology’s most significant benefits: • Stronger encryption methods ensure the safety of patient data in the health cloud. • To ensure data security and speed, the health cloud uses an EGC mechanism that is tuned for scalability and uploading speed. • It provides a high level of protection for data from insider threats. The rest of this paper is prearranged as follows. Section 2 delivers a comprehensive review of the relevant scholarly literature. With the help of an overall system design, the proposed methodology is clarified in Sect. 3. Section 4 details the proposed system’s performance evaluation, and Sect. 5 closes the article with recommendations for future research.

2 Related Works In order to protect against smart health threats, Zhang et al. implemented CP-ABE (Ciphertext Policy Attribute Based Encryption). Smart healthcare’s application of CP-ABE brings with it a unique set of challenges [16]. It was created to address these issues: a smart health access control system that takes privacy into account. In PASH, only the name attribute is made public, while the value of the access policy attribute is hidden in encrypted smart health records. In addition, attribute values

508

V. Gokula Krishnan et al.

typically contain more private information than other types. In this decryption test, PASH is able to successfully decrypt SHR (it requires few bilinear-pairings). Mobile Healthcare Social Networks are plagued by privacy concerns (MHSN). MHSN profile matching and data sharing are planned by Huang et al. [17] in the cloud. Identity Based Broadcast Encryption (IBBE) is used to outsource encrypted data to the cloud (IBBE). In addition, the doctor’s group receives data fast and safely. Using attribute-based conditional data re-encryption, the doctor’s referral is disseminated throughout the network to another doctor. A new enciphered text is generated from the encrypted one (without leaking the sensitive information). While integrating and exchanging E-health information, this book sought to address security and privacy concerns by providing a solution for Internet applications. Bao and colleagues [18] have presented a signal scrambling technique based on the application layer. To protect patient information, a minuscule amount of data is used to scramble the original. VOLU It uses either a random generator or a piece of data to derive the small data. Masood et al. [19] established a six-step architecture for measuring the patient’s physiological characteristics in Sensor Cloud Infrastructure (SCI). It begins with a preliminary selection, followed by an assessment of the patient’s physiological parameters and a security analysis. Finally, it estimates the functioning of the system. Cloud computing is a promising tool for healthcare data security. It’s a requirement, along with other security measures, while communicating electronically. Mbonihankuye et al.’s [20] leading strategy is the Health Insurance Probability and Accountability Act (HIPAA): [20]. Different analytical and conservational procedures can be used to ensure that healthcare data is properly recorded and kept. Data leaks and attacks on the cloud distributor may occur when the medical data is being published. The AFBS WOA algorithm, created by Thanga Revathi, et al. [21], combines AFBSO (Adaptive Fractional Brain Storm Optimization) with the Whale Optimization technique to address this issue [20, 21] (WOA). A new AFBS WOA algorithm generates the key matrices coefficients needed to retrieve a corrupted database and keep patient information private in the cloud. The secret key was calculated using a fitness function that incorporated utility and privacy considerations. A secure database can be built by multiplying the input database by a key matrix created by Tracy–Singh using the Tracy. Kumar et al. [22] extremely difficult to constantly monitor the central storage of health records that are vulnerable to security risks. For this reason, in order to protect confidential patient information, this study uses a block chain technology and a digital signature with authentication to protect it, as well as a cloud-based model to ensure the information’s authenticity and reliability. Traditional methods for preserving medical records were studied and compared to the model presented in the study, in terms of response time and the cost of storing and retrieving records. Smys [23] new technologies, such as sensor networks and smart monitors, have altered this picture by leveraging mobile devices and internet services. This has improved practical healthcare through predictive modeling and the acquisition of

Development of Secure Cloud-Based Healthcare Management Using …

509

more detailed individual measurements. A large amount of data allows researchers to analyse patterns [24] and trends in order to provide solutions that improve medical treatment while keeping costs down, while also ensuring that human lives are not put at risk. The survey on the accuracy and predictive power of big data analysis in the health care system is presented in this study.

3 Proposed System An outline of a way for safely transferring healthcare data between cloud systems is provided here.

3.1 Architecture Overview The following entities make up an efficient healthcare system based on EGC’s overall architecture (Fig. 1): HC: Users can store, update, and back up healthcare data using cloud services provided by the HC All cloud services are supported by the HC’s server, which houses all of the healthcare data. The health cloud’s data had to be protected from a variety of dangers. Encryption of data in the health cloud ensures the privacy of patient records.

Fig. 1 Architecture of optimized EGC-based secure health cloud

510

V. Gokula Krishnan et al.

TTP-CS: TTP-CS is the third-party-owned trusted entity that performs the cryptographic process outside of the cloud. The ECC algorithm has been modified in this mechanism’s design to ensure the security of sensitive healthcare records. In order to ensure that healthcare data may be shared in a secure manner, it is responsible for data confidentiality, integrity, key management. CUs: The health cloud’s clients are the cloud’s users (such as researchers, analysts, physicians, and others). Registration of CUs with the TTP-CS is required in order to carry out security services. Only one CU will own each data file, while all other CUs will be consumers of that information.

3.2 System Model For the safe transfer of medical data files to and from the cloud, this architecture supports asymmetric or public key cryptography. The TTP-CS receives a list of CUs and a patient health information file from the DO. It is as a result of this that TTP-CS generates two random 256-bit keys, the public key (K Pb) and private key (K Pr). An asymmetric key algorithm can be made to run for a shorter or longer period of time using several techniques. K Pb and K Pr are generated by using the SHA-256 hash function on a random number RN. For the encryption and decryption of healthcare data, it is used further. After the encryption or decryption procedure, no one has access to the complete key. For each CU, TTP-CS generates a unique K Pb that can be freely shared and used in the encryption process. However, K Pr is only known to the decryption unit and is not shared with any other units. Security objectives can be achieved by implementing these cryptographic operations.

3.2.1

Loading a File to HC

The TTP-CS receives the request for encryption when a CU wants to upload healthcare data to the health cloud. Access privileges are mentioned in the HDF and the cloud user’s list. Access to HDF may be Read-only or Read–Write depending on the CUL’s permissions for each CU. The TTP-CS creates the Access Control List (ACL) for healthcare data using CUL. The data owner will inform the novel CUL to the TTP-CS while the HDFs are being shared with a new center. Otherwise, it only transmits the center ID of the last remaining center in the chain of transmission. ACLs for each HDF are created and maintained by the TTP-CS once the encryption requisition has been received. The ACL is made up of data about files, such as the file’s ID, size, and owner ID, as well as metadata describing how the item was shared. The TTP-CS makes K Pr and K Pb for each CU after constructing the ACL. The HDF is then encrypted with an EGC-optimized encryption technique.

Development of Secure Cloud-Based Healthcare Management Using …

511

Elliptic Galois Cryptography Because of its foundation in elliptic curve theory, ECC is usually referred to as the public key encryption method. Instead of using conventional methods, the keys are created by utilizing the features of elliptic curve equations. EGC is employed in the proposed project. Elliptic curves over Galois fields (Fa) are used to improve calculation efficiency and eliminate rounding mistakes. It is possible to determine the Galois field’s value by utilizing the Mayfly algorithm’s best answer for the ideal value. Mayfly Algorithm The Mayfly method is used in this work to maximize CNN’s learning rate. To put it another way, Zervoudakis and his colleagues have presented a variation on PSO that incorporates the best of PSO, GA and FA. Because it has been demonstrated that PSO requires some modifications to ensure an optimal point when performing in high-dimensional spaces, researchers trying to improve the performance of the PSO algorithm using techniques like crossover and local search now have a powerful hybrid algorithmic structure based on the behavior of mayflies. A possible solution to the problem can be found by examining the mayfly’s location in the search space. The following is the flowchart for the algorithm. In the beginning, two groups of mayflies, one for each sex, are randomly formed. Each mayfly is randomly placed in the problem space as a potential solution and its performance is evaluated using the predetermined objective function, f, which is represented by the vector x =(x 1,…,x d) (x). The velocity of a mayfly is defined as the change in its position, and the flying direction of each mayfly is a dynamic interaction of individual and communal flying experiences. It is also possible for each mayfly to modify their trajectory toward their personal best (pbest) and the best position obtained by any swarm mayfly to date (gbest). (a) Movement of male mayflies It follows that the location of each male mayfly in a swarm is determined by both its own knowledge and that of its neighbors, as the males gather in swarms. (b) Movement of female mayflies Female mayflies do not form swarms, unlike their male counterparts. For the purpose of mating, they prefer to fly toward males. (c) Mating of mayflies Using the operator, two mayflies’ mating process is depicted as: One parent is chosen from the male, while the other is chosen from the female one. They attract each other in the same manner as parents attract their children. A random process or a fitness function can be used to make the selection. Likewise, the most beautiful woman is paired with the most attractive man. As a result of the mating, the following two children are born: o f f spring1 = L ∗ male + (1 − L) ∗ f emale

(1)

512

V. Gokula Krishnan et al.

o f f spring2 = L ∗ f emale + (1 − L) ∗ male

(2)

For example, L is a random variable that falls within a certain range for male and female parents. The starting velocities of offspring are set to 0 at the beginning of the game. The Mayfly Algorithm (MFA) can be stated in a pseudo code that shows the basic processes. Algorithm 1: Pseudo Code of MFA Objective f unction f (x), x = (x1, . . . , xd)T I nitiali ze the male may f ly population xi (i = 1, 2, . . . , N ) and velocities vmi I nitiali ze the f emale may f ly population yi (i = 1, 2, . . . , M) and velocities v f i Evaluate solutions Find global best gbest Do W hi l e stopping criteria ar e not met U pdate velocities and solutions o f males and f emales Evaluate solutions Rank the may f lies Mate the may f lies Evaluate the o f f spring Separate o f f spring to male and f emale randomly Replace wor st solutions with the best new ones U pdate pbest and gbest End whi l e Post − pr ocess r esults and visuali zation

The encrypted files as E f1 and f2 are the end result of this process. In this case, E f1 is the product of a random number k and a point on the elliptic curve Pt c that is randomly selected. HDF, k, and the public key, K Pb, are added together to form E f2. KPb is included into each CU’s ACL for the next step in the procedure. The integrity of each encrypted file is safeguarded by the HMAC signature and key generated and stored by the TTP-CS. This information is sent to the person who requested it: the center ID, the encrypted files (E f1 and E f2), and their K Pr. Whereas, just the center ID and K Pr are transmitted across a Secure Socket Layer to the rest of the CUs (SSL). After the encryption procedure, a secure overwrite separates K Pr and K Pb from the TTP-CS. It is up to DO or TTP-CS to upload the encrypted files (E f1 and E f2) after they have been received (on behalf of CU). The key generation procedure begins as soon as the encryption center is activated or the encrypted file is submitted. You have two options when it comes to uploading files: In either case, the DO can be promptly posted to the HC, as previously explained, or the TTP-CS can upload the file on behalf of the CU to the HC, which has the authority delegation. It is possible to upload a single file of medical data to the Health Center by following the steps listed below:

Development of Secure Cloud-Based Healthcare Management Using …

513

U1

The TTP-CS receives the healthcare data file from the physician and the list of users

U2

The ACL, private key, and public key for the physician are generated by TTP-CS. Next, an efficient EGC encryption mechanism is used to protect the data

U3

Physicians can get their TTP-CS private key, centre ID, and encrypted files by using the TTP-CS

U4

The physician upload the encrypted files straight to the health cloud

U5

TTP-CS uploads encrypted files to the health on behalf of the doctor in special cases

3.2.2

Downloading a File from the HC

It is either essential for the TTP-CS to receive an authentication request from the CU or for the DO to download encrypted files directly from the HC and then submit a decryption request. A locally upheld ACL authenticates the CU’s authorization from the HC. The TTPCS obtains K Pb from the ACL. The requesting CU will receive an access forbidden message if the ACL does not contain the K Pb. Because each CU has its own K Pb, no CU can use the K Pr of another CU. As a result, the decryption procedure can begin after the TTP-CS verifies the file’s integrity. Depending on whether or not the TTP-CS receives a valid K Pr, the decryption operation will either succeed or fail. After a successful ECC decryption, the HDF is sent to the relevant CU over an SSL connection. The secure overwriting approach eventually removes K Pr and K Pb from the TTP-CS. The TTP-CS can also be used to download files on behalf of the CU, same like the file uploading process. TTP-CS receives this request for decryption along with login credentials, as previously stated. Once the TTP-CS has confirmed that the CU for the specified file is genuine, it will forward this request on to the HC for processing. Further transmission of encrypted data takes place via TTP-CS, with the HC acting as a conduit. The rest of the process is the same as described previously. Here is an example of how to obtain a medical data file from the HC: D1

The TTP-CS receives requests from the CU

D2

In order for TTP-CS to send a download request to the health cloud, ACL verification is required

D3

The health cloud sends encrypted files to TTP-CS

D4

TTP-CS retrieves from the ACL

D5

It is sent to the appropriate CU with the original data file in it

514

3.2.3

V. Gokula Krishnan et al.

File Restore

ACL and key generation are not performed while recovering a file, unlike when uploading a file. TTP-CS receives a restore request from CUs (who have already downloaded the file) if any modifications have been made. True or false, TTP-CS verifies whether or whether CU has WRITE access to a file. The TTP-CS computes the keys if a valid request for file restoration has been made. Additionally, the file is encrypted before being subjected to the HMAC algorithm. They’re either reencrypted and transmitted through email or uploaded to the HCI server. Finally, the K Pr and K Pb are left out of the equation. The proposed model delivers the subsequent features to the healthcare data: Healthcare data must be protected from insider threats by preventing unauthorized access within the center. Secure sharing of healthcare data among the center.

3.2.4

Security Analysis

1. Eaves dropping The patient receives the private key from the certificate authority via a secure connection. As a result, hackers will be unable to access the encrypted data. 2. Replay attack The property and the secret key used to encrypt the files can be found in the tree structure. The EGC cryptography algorithms have been implemented by the doctors to their fullest potential. The optimization mechanism identifies the EGC values in order to identify the best solutions. As a result, a replay assault on the keyword provided by the patient is ruled out. Even if the hacker knows the characteristics and ciphertext, the secret key is not fixed in the EGC, therefore he can’t calculate it. 3. Masqurade and man in the middle attack (MIM) Hackers can’t use a masqurade attack because the properties are utilized to encrypt the files. The properties of the file must be known by the hacker if he wants to hack it. The hacker must know the property before he or she can change ciphertext files. Files are transmitted via cryptography rather than the MIM since we’ve utilized an EGC that operates on points instead than bytes.

4 Results and Discussion Using an Intel Core i5-6200U CPU clocked at 2.40 GHz and 8.00 GB of RAM, the proposed solution is put into practice on a Windows 10 64-bit OS system. The HC, TTP-CS, and CUs are the three main components described in the system model. Uses JPBC v.2.0.0 Java Pairing Based Cryptography library for communication between entities. Both elliptic curve and pairing procedures can be implemented with the

Development of Secure Cloud-Based Healthcare Management Using …

515

Table 1 Computation time for finding the value of EGC with different iterations No. of iterations

WOA

AFBS_WOA [21]

Mayfly

5

1.594

1.534

1.494

10

1.741

1.606

1.598

15

1.888

1.798

1.645

20

2.193

2.110

2.001

25

2.356

2.190

2.129

Table 2 Key generation time FS (MB)

Methodologies (time in second) AES

ECC

EGC

Proposed optimized EGC

10

1.594

1.534

0.004

0.00212

20

1.741

1.606

0.00425

0.00235

30

2.321

1.684

0.00476

0.00286

40

1.888

1.799

0.005

0.00302

50

1.952

1.866

0.00512

0.00328

60

2.193

1.923

0.0055

0.0035

70

2.286

2.034

0.00598

0.00398

80

2.694

2.129

0.00632

0.00427

90

2.827

2.388

0.00664

0.00463

100

2.887

2.545

0.00697

0.00499

help of its functions. It is possible to communicate between the entities thanks to the Java libraries. SSL encrypts all data sent and received. It was tested using the Cloudsim toolkit and evaluated in terms of key generation time, file upload and download times, and the time it took to discover EGC’s value. Performance analyses of proposed mayfly algorithm are tabulated on Tables 1, 2, 3 and 4.

4.1 Performance Analysis of Proposed Mayfly Algorithm See (Table 1).

4.2 Performance Analysis of Proposed Optimized EGC See (Tables 2, 3 and 4).

516

V. Gokula Krishnan et al.

Table 3 Time taken for uploading the encrypted files and downloading the decrypted files FS (MB)

Methodologies (time in second) AES

ECC

EGC

Proposed optimized EGC

UL

DL

UL

DL

UL

DL

UL

DL

0.1

1.4

0.99

1.48

1.15

0.80

0.80

0.70

0.70

0.5

1.48

1.03

1.89

1.31

0.94

0.96

0.80

0.82

1

2.06

1.48

2.90

1.85

1.24

1.18

1.20

1.24

10

14.95

9.90

14.59

10.45

6.43

6.48

5.60

5.68

50

58.56

35.57

60.37

35.90

9.01

10.24

8.25

8.78

100

112.41

59.14

115.15

61.59

17.39

20.68

16.35

18.98

500

492.03

229.81

872.09

400.21

33.24

39.25

31.10

38.22

Table 4 Speed of file uploading File size (MB)

Uploading speed (Mb/s)

0.1

11.5

0.5

12

1

11.9

10

12.92

50

12.5

100

13

250

13

500

13

5 Conclusion Increasing e-health productivity is now possible because to cloud computing-based health clouds, which allow medical professionals to access patient records from anywhere at any time, on any device. Secure data exchange between general practitioners, medical providers, and insurance companies is a critical concern for any healthcare company. In order to deal with this problem, encryption technologies are used to safeguard critical healthcare data. EGC-based encryption is utilized for data security in the proposed health cloud architecture. TTP-CS is also responsible for the encryption and decryption operations. Using the EGC model surpasses other existing systems in terms of key generation time, file upload time, file download time and uploading speed. EGC has a smaller key size, which makes key administration more simpler. The results demonstrate that EGC-based approach is a promising choice for safe healthcare data sharing in the health cloud. Because it hasn’t been developed to handle image-based data yet, this encryption approach can only be used to protect

Development of Secure Cloud-Based Healthcare Management Using …

517

plaintext. It is possible that this problem will be resolved in the future. Also implement the security mechanism in the cloud that are federated and then compare its efficiency with the existing methods.

References 1. Jayaraman I, Stanislaus Panneerselvam A (2021) A novel privacy preserving digital forensic readiness provable data possession technique for health care data in cloud. J Ambient Intell Humaniz Comput 12(5):4911–4924 2. Manne R, Kantheti SC (2021) Application of artificial intelligence in healthcare: chances and challenges. Curr J Appl Sci Technol 40(6):78–89 3. Qiu H, Qiu M, Liu M, Memmi G (2020) Secure health data sharing for medical cyber-physical systems for the healthcare 4.0. IEEE J Biomed Health Inform 24(9):2499–2505 4. Rushanan M, Rubin AD, Kune DF, Swanson CM (2014) Sok: security and privacy in implantable medical devices and body area networks. In: 2014 IEEE symposium on security and privacy. IEEE, pp 524–539 5. Sun Y, Lo FPW, Lo B (2019) Security and privacy for the internet of medical things enabled healthcare systems: a survey. IEEE Access 7(2019):183339–183355 6. Timothy DP, Santra AK (2017) A hybrid cryptography algorithm for cloud computing security. In: 2017 International conference on microelectronic devices, circuits and systems (ICMDCS). IEEE, pp 1–5 7. Banos O, Villalonga C, Damas M, Gloesekoetter P, Pomares H, Rojas I (2014) Physiodroid: combining wearable health sensors and mobile devices for a ubiquitous, continuous, and personal monitoring. Sci World J 8. Abdullah A, Ismael A, Rashid A, Abou-ElNour A, Tarique M (2015) Real time wireless health monitoring application using mobile devices. Int J of Comput Netw Commun (IJCNC) 7(3):13–30 9. Anuradha M, Jayasankar T, Prakash NB, Mohamed Yacin Sikkandar, Hemalakshmi GR, Bharatiraja C, Sagai Francis Britto A (2021) IoT enabled cancer prediction system to enhance the authentication and security using cloud computing. Microprocess Microsyst 80(2021):103301 10. Jaiswal K, Sobhanayak S, Turuk AK, Bibhudatta SL, Mohanta BK, Jena D (2018) An IoT-cloud based smart healthcare monitoring system using container based virtual environment in edge device. In: 2018 international conference on emerging trends and innovations in engineering and technological research (ICETIETR). IEEE, pp 1–7 11. Nirabi A, Hameed SA (2018) Mobile cloud computing for emergency healthcare model: Framework. Proc Int Conf Comput Commun Eng:375–379 12. Kumar V, Bhardwaj A (2020) Deploying cloud-based healthcare services: a holistic approach. Int J Serv Sci Manag Eng Technol (IJSSMET) 11(4):87–100 13. Blakley GR, Borosh I (1979) Rivest-Shamir-Adleman public key cryptosystems do not always conceal messages. Comput Math Appl 5(3):169–178 14. Gola KK, Gupta B, Iqbal Z (2014) Modified RSA digital signature scheme for data confidentiality. Int J Comput Appl 106(13) 15. Ali A, Pasha MF, Ali J, Fang OH, Masud M, Jurcut AD, Alzain MA (2022) Deep learning based homomorphic secure search-able encryption for keyword search in blockchain healthcare system: a novel approach to cryptography. Sensors 22(2):528 16. Zhang Y, Zheng D, Deng RH (2018) Security and privacy in smart health: efficient policy-hiding attribute-based access control. IEEE Internet Things J 5(3):2130–2145 17. Huang Q, Yue W, He Y, Yang Y (2018) Secure identity-based data sharing and profile matching for mobile healthcare social networks in cloud computing. IEEE Access 6:36584–36594

518

V. Gokula Krishnan et al.

18. Bao S-D, Chen M, Yang G-Z (2017) A method of signal scrambling to secure data storage for healthcare applications. IEEE J Biomed Health Inform 21(6):1487–1494 19. Masood I, Wang Y, Daud A, Aljohani NR, Dawood H (2018) Towards smart healthcare: patient data privacy and security in sensor-cloud infrastructure. Wireless Commun Mobile Comput 2018(Nov 2018), Art. no. 2143897 20. Mbonihankuye S, Nkunzimana A, Ndagijimana A (2019) Healthcare data security technology: HIPAA compliance. Wireless Commun Mobile Comput 2019(Oct 2019), Art. no. 1927495 21. Thanga Revathi S, Gayathri A, Kalaivani J, Christo MS, Pelusi D, Azees M (2021) Cloudassisted privacy-preserving method for healthcare using adaptive fractional brain storm integrated whale optimization algorithm. Secur Commun Netw 22. Kumar D, Smys S (2020) Enhancing security mechanisms for healthcare informatics using ubiquitous cloud. J Ubiquitous Comput Commun Technol 2(1):19–28 23. Smys S (2019) Survey on accuracy of predictive big data analytics in healthcare. J Inf Technol 1(02):77–86 24. Zervoudakis K, Tsafarakis S (2020) A mayfly optimization algorithm. Comput Ind Eng 145:106559

A Review of Mobile Computation Offloading Techniques M. Jyothirmai, Kesavan Gopal, and M. Sailaja

Abstract More and more people are increasingly using multimedia on their mobile devices, such as smartphones, tablet PCs and smart watches as a result of technological improvements. The most significant elements of a mobile device are the battery life, memory, bandwidth, and CPU performance. When such computationally extensive tasks are performed on a mobile device, the battery quickly drains. However, offloading such tasks to a proxy and executing those results in significant power savings in mobile devices. We compare the ways of offloading to a proxy from a mobile device in this study based on power usage, energy, and execution time. A thorough examination of the offloading process is also presented. The findings show a significant reduction in the amount of energy consumed by mobile devices. Keywords Mobile devices · Computation offloading · Task partitioning

1 Introduction Because of the growing popularity of using mobile devices e.g., smartphones, smart watches, and electronic tablets in people’s daily lives, demand for multimedia services (e.g., voice over IP (VoIP), web-browsing, video-watching, and filedownloading) is skyrocketing, resulting in an explosion in mobile traffic [1]. Multimedia apps that need a lot of resources, such as 3D video games, are becoming increasingly popular on mobile phones.

M. Jyothirmai (B) · M. Sailaja JNTUK, Kakinada, Andhra Pradesh, India e-mail: [email protected] K. Gopal LPU, Phagwara, Punjab, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_39

519

520

M. Jyothirmai et al.

Mobile devices will continue to develop even as their hardware and mobile networks evolve and improve. Always be resource-scarce, insecure, with erratic connectivity, and limited energy. Mobile applications are growing increasingly popular, and there has been a large increase in mobile subscriptions, indicating that mobile cloud applications and services are in high demand among mobile device users. However, because mobile terminals have limited resources, executing these applications directly on them does not meet the user’s expectations. As a result, computation is required on mobile devices and the performance and functionality are currently limited. To summarize, smartphones offer multi-core CPUs, better screen resolution, more memory, additional sensors, and radios, as well as a vast array of apps. These factors combined place a significant strain on the battery’s energy consumption [2]. Much research has been done to save energy to increase the life of batteries [2–5]. Computation offloading is a popular energy-saving strategy for mobile devices, in which applications make use of resource-rich infrastructures by transferring computation to these infrastructures. It was also discovered that transferring processing to the cloud via networks can help lower smartphone power usage. Mobile data offloading is a viable solution for addressing the aforementioned issues by leveraging complementary and novel networking approaches to transmit mobile data that was originally intended for cellular networks. Mobile data offloading is projected to play a significant role in mobile networks in the future, given the continued and rapid growth of mobile traffic. Offloading may help mobile devices conserve energy and function better. This, however, is typically dependent on several factors, including network bandwidths and data transfer rates. Various infrastructures and techniques have been proposed to improve offloading. They address issues such as user transparency, privacy, security, mobility, energy efficiency, and so on. The goal of this work is to introduce readers to compute offloading research for mobile devices. The motives, strategies, technology enablers, and designs for compute offloading are discussed in this study. This document addresses existing limitations and research directions for algorithmic mechanisms and associated infrastructures, as well as serves as a collective reference for them. The following is how the paper is structured: In Sect. 2, a brief history of enabling technologies is described. Section 3 outlines infrastructures and tools designed to solve the issues of offloading, as well as two objectives for offloading: lower execution time and conserve energy. Section 4 explains why offloading will become more significant in the future, and Sect. 5 gives the conclusion.

2 Mobile Task Offloading In this section, we first cover the basics of mobile cloud off-loading systems before going into the primary obstacles and issues that come with making off-loading decisions.

A Review of Mobile Computation Offloading Techniques

521

Fig. 1 Architecture of the offloading process

Figure 1 depicts an off-loading system that uses a proxy server to deliver multimedia material to a portable device. This system is made up of three components: content servers, proxy servers, and mobile devices [6]. A mobile or handheld device is any networked resource, such as a handheld (personal digital assistant or PDA), a gaming device, or a wireless security camera. Material servers keep track of multimedia and database content and respond to queries by delivering data (such as images) to clients. All communication between mobile devices and servers is relayed through proxy servers. Proxy servers are powerful servers that, among other things, can compress and decompress images, transcode video in real time, access/provide directory services, and deliver services based on a rule base. As a result, mobile devices negotiate security, service quality, and content delivery with proxy servers. In turn, the proxy servers make requests the image/video/data stream to content servers according to the needs of the user It’s worth noting that mobile devices can also be used to produce and deliver data to other mobile devices via proxy servers connected to the network. The abbreviations used are shown in Table 1. Table1 Abbreviations

PDA

Personal Digital Assistant

MEC

Multi-access Edge Computing

SPEA

Strength Pareto Evolutionary Algorithm

EFFORT

Energy Efficient Framework for Offload Communication

MCC

Mobile Cloud Computing

MCM

Mobile Cost Monitor

WLAN

Wireless LAN

522

M. Jyothirmai et al.

3 Major Criteria in Offloading Metrics: Most offloading decisions are based on a certain criterion. On one hand, energy, cost, and storage are criteria that should be minimized, while performance, robustness, and security are criteria that should be maximized [7]. Application Partitioning: Using the collected data, the offloading decision-making module makes a decision based on the metrics module (i.e., minimizing or maximizing some criteria), and then the partitioning module is called to divide the classes that make up an application into local and remote partitions, with the former running locally on the mobile device and the latter offloaded to a dedicated cloud server [8]. Load balancing: At any time, the edge node or datacenter may be overwhelmed, causing delays and, in some cases, inaccurate results [9]. Scalability: Real-time applications are frequently used by multiple users at the same time, which is managed by separate algorithms that are responsible for a large number of user requests. The offloading process, on the other hand, must be optimized in order for the application to scale up without causing harm [10]. Energy consumption trade-off: Because the offloading procedure consumes energy and bandwidth, the offloading decision must be made in light of this trade-off [10]. Availability: Mobile devices must always connect to the edge/cloud, which is difficult owing to network coverage gaps, congestion, limited bandwidth, and other networkrelated issues [10]. Security: Tasks and users’ data are transported via the network throughout the offloading process, increasing the risk of data theft and misuse. To address this problem, it is important to enlist the help of a third party [11]. Decision-making: Determining whether or not to offload a work is difficult due to a variety of factors such as delay, energy, and payment expenses [12]. Flexibility: When connecting to the edge, this poses a big issue [13]. Resource utilization: When compared to cloud infrastructure, the quantity of resources on the edge is limited. As a result, resource utilization management is essential to get the most out of limited resources while avoiding system overhead [14].

4 Classification Offloading can be divided into two categories: data offloading and computation offloading [15]. Data offloading is a technique for transferring data from a mobile device with limited storage and capability to a cloud repository. Data offloading from multi-mobile devices to multi-MEC servers was proposed by Zhang [16].

A Review of Mobile Computation Offloading Techniques

523

He proposed a game-centric pricing structure utilizing MEC to prioritize data offloading based on mobile device and resource allocation. Xu [17] proposed a timeefficient offloading strategy for intelligent sensors in edge computing, while maintaining privacy. An improved version of the Strength Pareto Evolutionary Algorithm (SPEA2) is utilized to jointly optimize average time consumption and average privacy entropy. Sun [18] constructed an optimization problem with weighting parameters in order to maximize the total efficiency of user computation. He also employed the iterative and gradient descent methods to demonstrate that the suggested strategy outperforms established methods. Computation offloading is a technique for processing intensive applications remotely in order to take advantage of strong resources in cloud servers, bypassing the CPU and battery limits on mobile devices [19]. Applications can run with reduced latency on mobile user equipment thanks to partial compute offloading [20]. Full and partial offloading are the most common offloading modes. The entire task is offloaded without partitioning in full offloading mode [21–24]. In partial offloading mode, the task is partitioned. As a result, some parts are processed locally while others are done remotely [20, 25–28]. By offloading compute-intensive operations to proxy servers, computation offloading is an effective technique to ensure user service quality [29]. The primary goal of computation is to shorten the service’s response time, increase the quality of the service Furthermore, when the mobile device doesn’t have the processing power, the computation can be moved to a proxy server or a cloud data center in order to increase the system’s overall performance. To do so many components of the computation offloading decision must be taken into account, such as maximizing performance and minimizing energy use.

5 Offloading Decision Before computation offloading, a number of questions must be answered. (1) When to offload? The task scheduler must be able to determine the offloading time slot for various scenarios and constraints. Because of the necessity for additional data transfers, which might increase time and/or energy consumption when task-related data is exchanged, remote execution on a cloud server is not always an advantageous option [8]. When the time and energy saved from offloading is insufficient to cover the additional communication expenses between the mobile device and the cloud, we can choose to run the app locally rather than offload it to the cloud. As a result, mobile cloud offloading is a viable option, but it is not required. We must determine the best moment to offload, such as when the wireless network is available, the amount of transmission data is low, or the amount of processing is high.

524

M. Jyothirmai et al.

Rahmati and Zhong [30] proposed a computation offloading method that allows smart mobile devices to dynamically adjust computational speeds based on calculation demands, lowering energy consumption and computation time. They improve the processing speed, transmission power, and offload rate on smart mobile devices to reduce energy consumption and application execution delays. For single-server and multi-server scenarios, they created a new compute offloading approach. There is a basic trade-off between mean energy usage and mean response time for diverse applications [31]. Offloading, in which cloud services are available with minimal network latencies, can provide better service by reducing response time. Miettinen and Nurminen [32] proposed offloading operation by switching between multiple transmission technologies, as well as examining the trade-off between energy usage for Wi-Fi search and transmission efficiency when the Wi-Fi network was accessible. Energy efficient delayed network selection has been used to optimize the trade-off between energy usage and data transmission delay by postponing data transmission until the device finds an energy-efficient network [33]. Some research was conducted to determine whether it is preferable to offload computational processes to a dedicated server rather than conducting them on the mobile device [34]. Mobile cloud offloading has the potential to reduce mobile device execution time and energy consumption, but the savings must outweigh the higher communication costs between the device and the cloud. [35]. Because the energy overhead of data transport may outweigh the energy savings from lower CPU usage, a performance seeking offload cannot guarantee energy savings [36]. When a wireless network is available, regardless of network quality, all traffic is promptly offloaded to the remote cloud [37]. When there isn’t a high-quality network available right now, the offloading procedure might be postponed until a suitable network becomes available [38]. We can opt not to offload the work to the cloud when the link is bad or the amount of data is huge, rather than transferring it directly to the cloud. (2) What to offload? If a task can be offloaded, the task scheduler must be able to determine whether it may be offloaded and, if so, what should be offloaded. Is it better to discharge in stages or all at once? Cloud data center workloads must be offloaded to edge devices and edge servers for edge computing to reduce service latency and network bandwidth utilization. As a result, workloads in cloud data centers must be partitioned, with some of them destined for edge devices and servers [39]. Cuervo et al. [40] recommended offloading and caching on the edge server instead of transferring the centralized database to the client to enhance application latency. Furthermore, filtering a high number of server requests through web proxies can reduce server workload dramatically. To accomplish a specific performance goal, such as the shortest reaction time or the lowest energy consumption, it should be established which portions of the task should be stored on the cloud server and which should be kept on the

A Review of Mobile Computation Offloading Techniques

525

mobile device [39]. Chen et al. [40] proposed a network caching technique that uses the storage capacity of multiple network devices to reduce network traffic. Some operations should always be performed locally on the mobile device, either because transferring relevant data would take too long and consume too much energy, or because these functions require access to local components (e.g., camera, GPS, user interface, accelerometer, or other sensors) [41]. To alleviate the strain of rapidly rising demand for caching and computing services, Lin et al. [42] proposed a new information-centric heterogeneous network design for content caching and computation. Malik [43] proposed that data-intensive edge computing applications be updated in the main database, eliminating the need for a copy to be sent to the server. They also provided a wide-area replication mechanism for delivering dynamic material while utilizing the capabilities of edge computing. (3) What is the optimum site to outsource workload execution and where should it be done? It’s critical to locate the greatest cloud service for offloading, i.e., the ideal spot to unload. By distributing partitioned tasks to mobile devices and proxy servers, task offloading can be done. The selection of targeted edge devices and edge servers involves the optimization of several objectives such as performance, energy, network bandwidth, and data privacy protection. To save energy, for example, energy-intensive activities are offloaded to cloud servers, and data-intensive jobs are offloaded to edge servers to reduce latency and network traffic [39]. In [44], an Energy efficient framework for offload communication (EFFORT) in MCC was used for communication offloading, resulting in a significant reduction in energy consumption. Flores and Srirama [45] devised a decision tree-based method for creating an offloading strategy in which all computation-intensive offloading decision operations are sent to a distant server for execution. An application’s components can be deployed on numerous application processing nodes, such as a mobile device, cloudlet, or cloud, implying that many offloading destinations and targets are possible [46]. In mobile cloud computing systems, Flores and Srirama [45] introduced a context-aware offloading method and a computation offloading approach. To estimate task execution costs, it supplied a general cost estimates model for cloud resources, including execution time and energy usage. A Mobile Cloud Middleware (MCM) was designed as a bridge between the mobile device and the cloud in [8], with the purpose of handling asynchronous delegation of mobile tasks to cloud resources and minimizing the time it takes to transfer duties from mobile devices to the cloud. In [47], they developed an adaptive offloading technique based on Lyapunov optimization that decides where each application activity should be performed so that energy consumption is minimized while latency is kept to a minimum. Effective offloading selections are made after assessing where offloading will increase system performance or generate the highest gains. (4) How to offload?

526

M. Jyothirmai et al.

Using one or more of the available wireless networks, mobile cloud offloading moves heavy computing from mobile devices to powerful cloud servers. Offloading work to a dedicated resource can be accomplished in a variety of ways, including using a cellular connection or a sporadic WLAN hotspot [48]. In most study papers [49, 50] it is assumed that cellular networks are always available to mobile users, whereas WiFi network availability varies by location to make decision-making easier. On their mobile devices, users move in and out of a WiFi coverage area. Chen et al. [51] suggested an online task offloading strategy to reduce the time it takes for mobile apps to run. They observed that load balancing methods might be used to dump jobs onto the cloud, maximizing parallelism between mobile and cloud for parallel tasks. Khan et al. [52] investigated the design of mobile edge computing computation offloading mechanisms in 5G heterogeneous networks and proposed the EECO framework as a multi-device energy-saving computation offloading framework to reduce computing task energy consumption during computation offloading and execution. Offloading decisions can be made in a fixed, static way, or offloading can be done dynamically in response to the application’s activity [58]. A good offloading choice is made by calculating the optimal time to offload under various device parameters, such as available bandwidth, amount of data to transfer, and energy consumption, and then picking the appropriate component to offload by separating a specific application into local and remote parts and determining the best place to offload under various cloud resource conditions.

6 Computation Offloading Resource-intensive calculations can be offloaded by shifting them to a separate processor such as a cluster, grid, or cloud. Offloading to a coprocessor can speed up a variety of tasks, including image watermarking, image compression, and mathematical computations. By offloading processing to an external platform across a network, it is feasible to get around a device’s physical limitations, such as its constrained computational power, storage space, and energy. Computation offloading model is shown in Fig. 2. Table 2 summarizes the previous work done in computation offloading. We present the contribution of the work done by using different partitioning algorithms with the goal of prolonging smartphone battery life or improving energy efficiency. The volume of mobile data is expected to expand fast, as seen in Fig. 3. Large volumes of multimedia data are generated by mobile platforms, and the majority of this material is kept online on cloud servers. When millions of cameras, microphones, GPS, and a variety of other sensors are connected, the amount of data generated is

A Review of Mobile Computation Offloading Techniques

527

Fig. 2 Computation offloading model

enormous. As the number of connected devices such as smartphones, tablets, laptops and sensors increases, so will the demand for more features. These factors suggest that mobile computing rates will not keep pace with the growth of data and the processing demands of applications. On one hand, there has been a significant increase in the types and amounts of mobile data, as well as the computational requirements of mobile applications. On the other hand, the computing capabilities of the devices, those that receive and store data and deliver user applications—are unlikely to keep up. Computation offloading is a natural answer to this issue.

7 Offloading Techniques The major issue with smartphones that needs to be fixed is their high battery consumption. By shifting mobile applications to the cloud and solving the issues there, the

528

M. Jyothirmai et al.

Table 2 Literature review Year Authors

Contribution

2020 Elgendy [21]

For mobile-edge computing, an efficient and secure multi-user multi-task compute offloading model with guaranteed performance in latency, energy, and security

2019 Nguyen [[9]

Maximize computation offloading and resource allocation to reduce the weighted sum of all mobile users’ energy consumption

2019 Sun [18]

A combined computation approach that combines two schemes, local computing and data offloading

2019 Kuang [25]

For MEC systems with several separate jobs, there is a dual problem of partial offloading scheduling and resource allocation

2018 Ning [20]

An Iterative Heuristic MEC Resource Allocation (IHRA) algorithm is used to make a dynamic offloading decision

2018 Akherfi [13]

Highlights current offloading frameworks and computation offloading strategies, as well as their most important challenges, and assesses them

2015 Tao and Yaling [3] They attempt to determine which components of a mobile application should be offloaded to the cloud for a given set of computational components 2015 Wu [8]

Offloading computation-intensive elements of mobile apps to a capable cloud server is an effective technique to reduce a resource-constrained mobile device’s struggle with resource-hungry mobile apps, hence improving the device’s performance

2014 Xia and Feng [2]

Reduces the execution time of an application running on a smartphone by offloading computing to the cloud. This improves the energy economy of smartphones and improves the application’s performance

2014 Zhang [5]

Offloading service that allows a mobile application to easily offload some of its work from mobile terminals to the cloud

2013 Wu [7]

Based on the tradeoff between reducing execution time and extending the battery life of mobile devices, a novel adaptive offloading strategy is presented and examined

2012 Kovachev [1]

Allows for the expansion of android application execution from a mobile client to the cloud in an adaptable manner

2006 Kejriwal [6]

A method of partitioning the watermarking, embedding, and extraction algorithms and offloading some tasks to a proxy server

energy consumption issue could be solved. This section explains some computational offloading models. MCC Mobile cloud computing technique uses a cloud-based copy of the communication method. A message is delivered from the program to the cloud-based clone, which then sent it to the service. That service has two ends, one of which is on the cloud and the other on a mobile device. The cloud service pulls the data from the cloud copy and then sends it to the smartphone. The cloud-based service recognizes the

A Review of Mobile Computation Offloading Techniques

529

Fig. 3 Growth in mobile traffic

communication with the mobile device. The mobile device selects the appropriate application and sends the data to it. As a result, there will be only one service running on mobile devices rather than a service for each application. Clone Cloud The Clone Cloud framework was developed with the goal of extending the battery life and enhancing performance on mobile devices by offloading resource-intensive components to cloud servers. In this approach, the partitioning step combines static program analysis with program profiling to create a set of off loadable components while adhering to certain requirements, such the requirement that methods that employ mobile sensors should be run locally. MEC In mobile edge computing, compute-intensive operations are offloaded either so that mobile devices can use less power or because they cannot be completed in time due to hardware limitations. MEC servers allow for the consideration of novel applications as possibilities for offloading, resulting in bandwidth savings and scalability. MAUI In order to calculate the energy consumption during code execution based on the number of CPU cycles needed to run it. In MAUI the mobile device is profiled and a straightforward linear model is created.

8 Conclusion We review the existing work in the subject of mobile compute offloading in this paper. We also go through the current frameworks for computation offloading, as well as

530

M. Jyothirmai et al.

the many strategies utilized to improve the capabilities of smartphone devices using cloud resources. We’ve noticed that current offloading frameworks are still dealing with various issues. We feel that pursuing other options, such as adopting a middleware-based design with an optimal offloading mechanism, could improve the frameworks already available and deliver more efficient and adaptable solutions. Offloading strategies that take advantage of cloud computing will become increasingly relevant as various linked devices become more widely deployed. Applications on these linked devices will begin to be created with offloadable computing in mind, and this type of application design will benefit from the methodologies and solutions discussed in this paper. Hence we conclude that computation offloading can be used to reduce energy consumption in mobile devices. However, this technique has its own limitations. In order to achieve maximum efficiency, the device should be able to perform certain tasks at low power levels. This means that the device needs to be able to run at lower clock speeds than normal. Another limitation is that the device cannot use multiple cores simultaneously.

References 1. Kovachev D, Yu T, Klamma R (2012) Adaptive computation offloading from mobile devices into the cloud. In: 2012 IEEE 10th international symposium on parallel and distributed processing with applications. IEEE 2. Xia F et al. (2014) Phone2Cloud: exploiting computation offloading for energy saving on smartphones in mobile cloud computing. Inf Syst Front 16(1):95–111. Elissa K, Title of paper if known, unpublished 3. Tao Y, Zhang Y, Ji Y (2015) Efficient computation offloading strategy for mobile cloud computing. Proc IEEE Int Conf Adv Inf Net Appl (AINA 2015) 4. Elgazzar K, Martin P, Hassanein HS (1989) Cloud-assisted computation offloading to support mobile services. IEEE Trans Cloud Comput 4(3):279–292. Young M (1989) The technical writer’s handbook. University Science, Mill Valley, CA 5. Zhang Z, Lim H, Lee HJ (2014) An efficient framework for computation offloading in mobile cloud computing. Int Conf Future Inf Commun Eng 6(1) 6. Kejariwal A et al. Energy efficient watermarking on mobile devices using proxy-based partitioning. IEEE Trans Very Large Scale Integr (VLSI) Syst 14(6):625–636 7. Wu H, Wang Q, Wolter K (2013) Tradeoff between performance improvement and energy saving in mobile cloud offloading systems. Proc Int Conf Commun Workshops (ICC):728732 8. Wu H (2015) Analysis of offloading decision making in mobile cloud computing. Ph.D. dissertation, Department of FB Mathematik und Informatik, Freie Universitaet, Berlin, Germany 9. Nguyen PD, Ha VN, Le LB (2019) Computation offloading and resource allocation for backhaul limited cooperative MEC systems. In: Proceedings of the 90th vehicular technology conference (VTC2019-Fall), pp 1–6 10. Aazam M, Zeadally S, Harras KA (2018) Offloading in fog computing for IoT: review, enabling technologies, and research opportunities. Future Gener Comput Syst 87:278–289 11. Wang F, Diao B, Sun T, Xu Y (2020) Data security and privacy challenges of computing offloading in FINs. IEEE Network 34(2):14–20

A Review of Mobile Computation Offloading Techniques

531

12. Behera SR, Panigrahi N, Bhoi S, Sahani A, Mohanty J., Sahoo D, Maharana A, Kanta LP, Mishra P (2020) A novel decision making strategy for computation offloading in mobile edge computing. In Proceedings of 2020 international conference on computer science, engineering and applications (ICCSEA), pp 1–5 13. Akherfi K, Gerndt M, Harroud H (2018) Mobile cloud computing for computation offloading: Issues and challenges. Appl Comput Informat 14(1):1–16 14. Peng K, Zhao B, Xue S, Huang Q (2020) Energy- and resource-aware computation offloading for complex tasks in edge environment. Complexity 15. Alqarni MM, Cherif A, Alkayal E (2021) A survey of computational offloading in cloud/edgebased architectures: strategies, optimization models and challenges. KSII Trans Internet Inf Syst (TIIS) 15(3):952–973 16. Zhang T (2017) Data offloading in mobile edge computing: a coalition and pricing based approach. IEEE Access 6:2760–2767 17. Xu Z et al. (2019) A time-efficient data offloading method with privacy preservation for intelligent sensors in edge computing. EURASIP J Wireless Commun Netw 2019(1):1–12 18. Sun H, Zhou F, Hu RQ (2019) Joint offloading and computation energy efficiency maximization in a mobile edge computing system. IEEE Trans Veh Technol 68(3):3052–3056 19. Khan MA (2015) A survey of computation offloading strategies for performance improvement of applications running on mobile devices. J Netw Comput Appl 56:28–40 20. Ning Z et al. (2018) A cooperative partial computation offloading scheme for mobile edge computing enabled Internet of Things. IEEE Internet Things J 6(3):4804–4814 21. Elgendy IA et al. (2020) Efficient and secure multi-user multi-task computation offloading for mobile-edge computing in mobile IoT networks. IEEE Trans Netw Serv Manag 17(4):2410– 2422 22. Guo H, Liu J, Qin H (2018) Collaborative mobile edge computation offloading for IoT over fiber-wireless networks. IEEE Network 32(1):66–71 23. Valentino R, Jung W-S, Ko Y-B (2018) A design and simulation of the opportunistic computation offloading with learning-based prediction for unmanned aerial vehicle (uav) clustering networks. Sensors 18(11):3751 24. Luo J et al. (2019) QoE-driven computation offloading for edge computing. J Syst Architect 97:34–39 25. Kuang Z et al. (2019) Partial offloading scheduling and power allocation for mobile edge computing systems. IEEE Internet Things J 6(4):6774–6785 26. Saleem U et al. (2020) Latency minimization for D2D-enabled partial computation offloading in mobile edge computing. IEEE Trans Veh Technol 69(4):4472–4486 27. Qin M et al. (2017) Service-oriented energy-latency tradeoff for iot task partial offloading in mec-enhanced multi-rat networks. IEEE Internet Things J 8(3):1896–1907. Mach P, Becvar Z (2017) Mobile edge computing: a survey on architecture and computation offloading. IEEE Commun Surv Tuts 19(3):1628–1656, 3rd Quart 28. Nir MPS (2014) Scalable resource augmentation for mobile devices. Ph.D. dissertation, Department of Systems and Computers Engineering, Carleton University, Ottawa, ON, Canada 29. Wang Y, Sheng M, Wang X, Wang L, Li J (2016) Mobile-edge computing: partial computation offloading using dynamic voltage scaling. IEEE Trans Commun 64(10):4268–4282 30. Rahmati A, Zhong L (2007) Context-for-wireless: Context-sensitive energy-efficient wireless data transfer. In: Proceedings of 5th international conference mobile systems applications services, pp 165–178 31. Lin Y-D, Chu ET-H, Lai Y-C, Huang T-J (2013) Time-and-energyaware computation offloading in handheld devices to coprocessors and clouds. IEEE Syst J 9(2):393405 32. Miettinen AP, Nurminen JK (2010) Energy efficiency of mobile clients in cloud computing. In: Proceedings of 2nd USENIX conference hot topics in cloud computing, p 4 33. Gember A, Dragga C, Akella A (2012) ECOS: practical mobile application offloading for enterprises. In: Proceedings of international conference mobile systems application services, p4

532

M. Jyothirmai et al.

34. Mehmeti F, Spyropoulos T (2013) Performance analysis of ‘on-the-spot’ mobile data offloading. In: Proceedings global communication conference (GLOBECOM), pp 1577–1583 35. Mehmeti F, Spyropoulos T (2014) Is it worth to be patient? Analysis and optimization of delayed mobile data offloading. In: Proceedings INFOCOM:2364–2372 36. Jiang C et al. (2019) Toward computation offloading in edge computing: a survey. IEEE Access 7(2019):131543–131558 37. Yuan C, Chen Y, Zhang Z (2004) Evaluation of edge caching/offloading for dynamic content delivery. IEEE Trans Knowl Data Eng 16(11):1411–1423 38. Wu H, Knottenbelt W, Wolter K, Sun Y (2016) An optimal offloading partitioning algorithm in mobile cloud computing. In: Proceedings international conference quantitative evaluation systems, pp 311–328 39. Chen M, Hao Y, Qiu M, Song J, Wu D, Humar I (2016) Mobility-aware caching and computation offloading in 5G ultra-dense cellular networks. Sensors 16(7):974 40. Cuervo E et al (2010) MAUI: making smartphones last longer with code offload. In: Proceedings of the 8th international conference mobile systems applications services, pp 49–62 41. Zhou, Yu FR, Chen J, Kuo Y (2017) Resource allocation for information-centric virtualized heterogeneous networks with in-network caching and mobile edge computing. IEEE Trans Veh Technol 66(12):11339–11351 42. Lin Y, Kemme B, Patino-Martinez M, Jimenez-Peris R (2007) Enhancing edge computing with database replication. In: Proceedings 26th IEEE international symposium reliable distributed systems (SRDS):45–54 43. Malik SUR et al. (2021) EFFORT: energy efficient framework for offload communication in mobile cloud computing. Softw Pract Experience 51(9):1896–1909 44. Wu H, Huang D (2014) Modeling multi-factor multi-site risk-based offloading for mobile cloud computing. In: Proceedings 10th international conference networks service management (CNSM), pp 230–235 45. Flores H, Srirama SN (2014) Mobile cloud middleware. J Syst Softw 92:8294 46. Wang C, Li Z (2004) A computation offloading scheme on handheld devices. J Parallel Distrib Comput 64(6):740–746 47. Hyytiä E, Spyropoulos T, Ott J (2015) Offload (only) the right jobs: Robust offloading using the Markov decision processes. In: Proceedings 16th international symposium world wireless, mobile multimedia networks (WoWMoM), pp 19 48. Kim Y, Lee K, Shroff NB (2014) An analytical framework to characterize the efficiency and delay in a mobile data offloading system. In: Proceedings of the 15th international symposium on mobile ad hoc networking computing, pp 267–276 49. Mehmeti F, Spyropoulos T (2016) Stay or switch?: Analysis and comparison of delays in cognitive radio networks with interweave and underlay spectrum access. In Proceedings 14th ACM international symposium mobility management wireless access (MobiWac), pp 9–18 50. Jia M, Cao J, Yang L (2014) Heuristic offloading of concurrent tasks for computationintensive applications in mobile cloud computing. In: Proceedings IEEE conference computing communication workshops (INFOCOM WKSHPS), Toronto, ON, Canada, pp 352–357 51. Chen X, Jiao L, Li W, Fu X (2016) Efficient multi-user computation offloading for mobile-edge cloud computing. IEEE/ACM Trans Netw 24(5):2795–2808 52. Khan MA (2015) A survey of computation offloading strategies for performance improvement of applications running on mobile devices. J Netw Comput Appl 56:28–40

Study of the Impact of Sybil Attack in VANETs Using F2MD T. Pavithra, B. S. Nagabhushana, and Suchismitha Das

Abstract According to a survey, 137,726 accidents have occurred in the year 2018 alone. In order to avoid such accidents and also to provide other services, research on Vehicular Adhoc networks started. VANETs have hardware called OnBoardUnits (OBU) situated on vehicles and Fixed infrastructure situated on the Roadside called RoadSideUnits (RSU). Standards have been developed that allows communication between vehicle to vehicle and Vehicles to RSUs. Also, separate Bandwidth standard has been defined called Dedicated Short Range Communication (DSRC) to provide communication during emergency. However, we are yet to identify the associated risk of privacy and security in such communication. Then only successful implementation of Intelligent Transportation System can be achieved. Hence study of various attacks and its impact on Vehicular network is very much important. In this paper, we will perform the simulation of VANET and study the various performance metrics by introducing Sybil attack and analyse its impact over the network. Keywords Vehicular adhoc NETwork (VANET) · Framework for misbehavior detection (F2MD) · SUMO (simulation of urban mobility) · TraCI (traffic control interface) · ITS (intelligent transportation system) · MBD (misbehavior detection) · Vehicles in network simulation (Veins) · Objective modular network testbed (Omnet++ )

T. Pavithra (B) Department of Electronics and Communication Engineering Government, SKSJT Institute, Bangalore, India e-mail: [email protected] B. S. Nagabhushana Department of Electronics and Communication Engineering, BMS College of Engineering, Bangalore, India S. Das Tata Consultancy Services, Bangalore, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_40

533

534

T. Pavithra et al.

1 Introduction According to a recent survey, India reports maximum number of road accidents globally, with almost 1.5 lakh people getting killed and more than 4.5 lakh people getting into minor injuries, annually. The biggest reason for accidents is found to be the speed. One of the main goals of VANET is to ensure safety of the users. VANET is a type of MANETS (mobile Adhoc networks) in which vehicles are the mobile nodes which are served by the Road Side Units (RSUs) which act as the Access Points (APs). Each vehicular nodes can freely move in the network and nodes can communicate with each other with single-hop or multi-hop [2]. The main information used by the driver from VANET for guidance are traffic information, road information, and collision [1]. With proper interpretation and post processing messages, drivers become aware of the situation and the can take necessary decisions to prevent road accidents. At the same time, there is a possibility of inducing false messages or network may be flooded with false message to make the network unavailable to the genuine users by inducing some attacks in the network. Therefore, it is necessary to identify attacks and eliminate such attacking nodes from the network. In this paper we discussed a type of attack called Sybil attack. The Sybil attack is named after the book Sybil, a case study of a woman with multiple personality disorder. Sybil attack is an impersonation attack and in this type of attack, an attacker node creates an illusion among other nodes of a network as a set of nodes by taking false identities, or by creating new identities at the worst case. It is very easy to execute such attacks in adhoc network environment of which VANETs is a representative case. Here, an attacker node will send messages with different pseudonyms to other nodes present in the network. The nodes sending spoofed messages to other nodes are attacking nodes and nodes receiving the spoofed information are called Sybil nodes. All other types of attacks get easily emulated in the presence of Sybil attack. Such cases may just be creating an illusion of a traffic jam by using some fabricated non-existent nodes or an illusion of accident happened so that other vehicles can change their route [3]. The detection of Sybil attacks is quite difficult in VANETs. A node that is participating actively in VANET can take a fake identity and cause network disruption. Such an action is called misbehaviour in VANETs and it creates a safety issue in the network. Misbehaviour Detection (MBD) in Vehicular networks is a field of research which focus on detection of intrusion of nodes that causes trouble to the transportation safety. The Algorithms developed to counter MBD could either be sensor-based or Vehicle to Vehicle and Vehicle to RSU based. Hence, it is necessary to simulate the VANETs and evaluate the developed MBD algorithms using simulation results [4]. The work presented herewith is a contribution in this specific space.

Study of the Impact of Sybil Attack in VANETs Using F2MD

535

2 Literature Review on Sybil Attack Sybil attack was first introduced by Douceur [5] during his study on security issues in peer-to-peer networks. Then, Karlof and Wanger proved that Sybil attack cab pose a serious threat to network routing in case of Wireless Sensor Networks [6]. Grover et al. have analysed the effect of the Sybil attack on the performance of VANET by measuring the packet delivery ratio, throughput, packet drop and number of collisions with variation in node mobility by performing simulation. They executed different types of Sybil attacks over the vehicular network and the results obtained after simulating the network showed that the Quality of Service provided by the Vehicular network degrades if Sybil attack happens in the network and hence study of Sybil attack is important in VANETs [7]. They have proposed an approach for distributed Sybil attack, in which, nodes exchange the observed details with their neighbours cooperatively and help RSU in the detection of attack. Kanwalpreet Singh et al. have proposed an interesting technique for detection of Sybil attack. This technique makes use of signal strength and vehicles need to be in monitor mode. In this technique, the RSU will flood the ICMP messages through the network. The vehicles will send their signal strength to their nearest RSU while receiving the ICMP messages. The RSU will collect the entire information from the vehicles and exchange this data with other RSUs. If multiple signal strength is received for a specific node, it will be considered as the malicious node in the networks. In order to get a confirmation about the malicious node, the RSUs will send the control packets in the network. After receiving the control packets, vehicles will enter into the monitor mode and start watching their adjacent nodes. Once the malicious node is sensed, the technique of multiple path routing is applied to isolate the malicious nodes from the network. In [8], authors arrived at a conclusion that broadcasting technique is very efficient to select an efficient route from source to the destination. Due to decentralized nature of the network, it is quite easy for a malicious node to join the network. These nodes are responsible for various types of active and passive attacks [9–11]. One of the commonly used mobility and network simulators is Vehicle Network Simulator: VEINS. VEINS is a combination of Objective Modular Network Testbed in C++: OMNET++ and Simulation of Urban Mobility: SUMO. The simulator will pair an OMNeT ++ node for each vehicle present in the simulation and then pairs node movements with movements of vehicles in the road traffic simulator (i.e., SUMO). Then, both the network simulator and mobility generator can run in parallel[13]. This is possible due to a bidirectional coupling by a standardized connection protocol, called the Traffic Control Interface (TraCI). Unlike to NS2, it does not require any Vehicular mobility generator separately and TraCI helps in communication between Omnet++ and SUMO.

536

T. Pavithra et al.

One of the well-known simulation frameworks which allows us to evaluate the efficiency of Misbehavior Detection algorithms in Vehicular networks is Vehicular Reference Misbehaviour (VeReMi). It is an extension of VEINS and it has a dataset that consists of transmission data from both misbehaving nodes and genuine nodes. This can be used for the evaluation of Misbehaviour Detection algorithms offline. VeReMi also provides five attacks based on position and four algorithms for Misbehaviour Detection. In [4], authors have proposed simulation framework named Framework for Misbehaviour Detection: F2MD based on VEINS. F2MD allows performing the studies on a large set of vehicles and can be simulated using a large road network. F2MD can perform Local Misbehaviour Detection running a Misbehaviour Detection system by each entity to cope with insider attackers. In our paper, we are using veins-F2MD to evaluate the effect of Sybil attack on VANET and local detection of Sybil attack using MBD algorithm.

3 Sybıl Attack Sybil attack is a type of network attack where the attacker creates large number of fake identities associated with a single node to give an illusion of large number of nodes in a network[15]. These fake identities are called Pseudonyms. An attacker uses these pseudonyms to perform an attack such as creating an illusion of congestion in the network. There are multiple ways to create the fake identities for a sybil node. Also, the type of sybil attacks are different such as distributed storage, routing and data summarization. A type of Sybil attack includes generating the vehicles called ghost vehicles that will be used to perform attack. The sequence of actions will be used to perform the sybil attack include: • Here, the speed, direction and Position of the ghost vehicle will be decided based on the target vehicle. • Many number of pseudonyms will be generated and maintained. One pseudonym will beused for each ghost vehicle. • Attacker vehicle’s beaconing frequency will be increased based on the number of ghost vehicles.

4 Attacker Model In order to perform the simulation, we are making the following assumptions. Insider Attacker: The attacker will possess the required credentials to communicate in a C-ITS. Therefore, any node in the network can be converted to a malicious node.

Study of the Impact of Sybil Attack in VANETs Using F2MD

537

Active: Here, the attacking node will participate actively in C-ITS communication and sends beaconing data. He can even modify the data as he can access the data traversing through the network. Pseudonym certificates access: Here, the assumption is made that, the attacker has full access to Pseudonym certificate.

5 Sımulatıon 5.1 Evaluation Veins is built on: OMNET++ (discrete event simulator) and SUMO (traffic simulator). SUMO will provide the traffic patterns for the realistic map and OMNET++ will provide different modules such as Application layer, Dedicated Short Range Communication, and a Physical layer to provide the realistic network behaviour. A small patch called Traffic Control Interface (TraCI) is used to establish a connection between OMNET++ and SUMO. To evaluate the effect of Sybil attack in VANETs, we created a map using OSM Web Wizard in SUMO as shown in Fig. 1. Firstly, we have introduced random vehicles in the default map of the veins network. When nodes are moving through the network, accident event is introduced at a random Location and time. The nodes send the event details to their neighbouring

Fig. 1 Map created using OSM web wizard in SUMO

538

T. Pavithra et al.

Fig. 2 Vehicles diverting routes to avoid congestion

nodes through broadcasting. Upon reception of this information, vehicles wait for some time and divert to some other route so that they can avoid the congestion as shown in Fig. 2. Then, we injected the malicious nodes by 10%, 30%, 50%, and 70% the network to understand the effect caused by such attackers over the network. The below table 1 shows the simulation parameters and its corresponding values.

5.2 Performance Evaluation Metrics For the evaluation of the performance over Vehicular networks in presence of Sybil attack, the following metrics have been considered [12].

Study of the Impact of Sybil Attack in VANETs Using F2MD Table 1 Simulation setup parameters

5.2.1

539

Attributes

Value

Network simulator

Omnet++ 5.4.1

Traffic simulator

Sumo 1.0.1

Framework V2x simulator

Veins-F2md

No. of vehicles

–

No. of malicious nodes

0, 10, 30, 50, 70%

No. of RSU

09

Simulation area

8 × 8 km

Communication range

1200 m

Simulation time

50 s

Content Delivery Ratio (CDR)

Content refers to the information that a vehicle want o convey.Since, VANETs have been developed with the interest of avoiding accidents by sending the sensitive data, delivery of the information to the intend vehicle is very important. Hence, the calculation of Content Delivery Ratio gives the major information about how efficiently data is sent/received over a network. CDR indicates the number of messages that are successfully received by the genuine vehicles. Let, MR represent the the number of messages received. MPRE represent the number of messages that are expected to be received in the network, CDR can now be calculated as: CDR = MR/MPRE We require this ratio to be high to assume a network to be efficient.

5.2.2

Packet Loss Ratio (PLR)

When the CDR is low, it indicates that network is not receiving as much information as needed which clearly says that some data is leaking. Therefor we need to calculate the amount of information lost. PLR gives the information about the number of messages that are lost due to the misbehaviour action of malicious nodes. If MT represents the total number of messages and out of MT, if ML number of messages are dropped in the network, then PLR can be calculated as: PLR = ML/MT

540

T. Pavithra et al.

In general, MT indicate the messages received by both genuine and malicious nodes. If MR represents the number of messages received by genuine nodes and ML represents the number of messages dropped at the Malicious nodes, then the total number of messages MT can be written as: MT = MR + ML In result section we are going to find how these paraameters get affected with increase in Sybil attack.

6 Results In this section we will see how CDR and PLR values vary with variation in the attacking vehicle density. As we can see in Fig. 3, the content delivery ratio is reduced with an increase in Attack Percentage.CDR was 100% without any attack and gradually reduces with increase in the percentage of attack. Therefore, Content Delivery Ratio can be considered as one of the parameter to identify that attack is being introduced in the network. Packet Loss Ratio, as shown in Fig. 4 increases and False message also increases as shown in Fig. 5 with an increase in attack percentage.

Fig. 3 Attack percentage versus content delivery ratio

Study of the Impact of Sybil Attack in VANETs Using F2MD

541

Fig. 4 Attack percentage versus packet loss ratio

Fig. 5 Attack percentage versus false message

7 Conclusion From our simulation result, it is clear that sybil attack affects the network performance by reducing the Content Delivery Ratio and by incresing the Packet loss and False

542

T. Pavithra et al.

messages. But the security in VANETs is very much important. Many researcher are working on these issues to learn how security can be provided in VANETs. Author of [14] proposed a method involving cloud computing to improve the network performance and The approximate member query filter is utilised to avoid interaction between attackers. Inorder to avoid VANET attcks so as to keep the performance of the network high, we need to detect the attack as soon as it is introduced in the network. This requires developing a trained model which can detect the and in turn, developing a trained model requires too much of data both with network attack and without any attack in the network. Therefore, In our next paper, we will use the available data from different research institutes and learn how different Machine Learning models help us in identifying the attack and which model gives better classification accuracy in identifying the attack.

References 1. Jain M, Saxena R (2017) Overview of VANET: requirements and its routing protocols. In: International conference on communication and signal processing, India, pp 6–8 2. Samara G, Al-Salihy WAH, Sures R (2011) Security analysis of vehicular Ad-Hoc networks (VANET), National advanced IPv6 Center, Universiti Sains Malaysia, Penang, Malaysia; *School of Computer Science, Universiti Sains Malaysia, Penang, Malaysia. In: 2010 Second ınternational conference on network applications, protocols and services. network security & ıts applications (UNSA), vol 3, no 6 3. Rahbari M, Jamali MAJ (2011) Effıcıent detectıon of sybıl attack based on cryptography ın VANET, Department of Computer Science, Shabestar Branch, Islamic Azad University, Shabestar, Iran. Int J Netw Secur Appl (UNSA) 3(6) 4. Kamel J, Ansari MR, Petit J, Kaiser A, Jemaa IB, Urien P (2013) Simulation framework for misbehavior detection in vehicular networks. IRT SystemX, Palaiseau, France; Telecom Paris Tech, Paris, France; # OnBoard Security , Inc., Wilmington, USA 5. Patil HK, Chen TM Chapter 18—Wireless sensor network security: the ınternet of things, pp 317–337 6. Karlof C, Wagner D (2003) Secure Routing in Wireless Sensor Networks: Attacks and Countermeasures. Ad Hoc Netw 1:293–315 7. Grover J, Kumar D, Sargurunathan M, Gaur MS, Laxmi V (2010) Performance evaluation and detection of Sybil attacks in vehicular Ad-Hoc networks. Commun Comput Inf Sci 89:473–482 8. Kanwalprit Singh, Harmanpreet Kaur, “Evaluation of Technique for detection of Sybil Attack on VANET” 9. https://veins.car2x.org/ 10. https://doc.omnetpp.org/omnetpp/manual/ 11. https://sumo.dlr.de/docs/ 12. Sommer C, German R, Dressler F (2011) Bidirectionally coupled network and road traffic simulation for ımproved IVC analysis. IEEE Trans Mob Comput (TMC) 10(1):3–15 13. Weber J, Neves M, Ferreto T (2021) VANET simulators: an updated review. J Braz Comput Soc 27:8. https://doi.org/10.1186/s13173-021-00113-x 14. Neelaveni R (2019) Performance enhancement and security assistance for VANET using cloud computing. J Trends Comput Sci Smart Technol 1(1):36–45. https://doi.org/10.36548/jtcsst. 2019.1.004

Study of the Impact of Sybil Attack in VANETs Using F2MD

543

15. Pavithra T, Nagabhushana BS (2020) A survey on security in VANETs. In: 2020 Second ınternational conference on ınventive research in computing applications (ICIRCA), pp 881– 889

Aatmanirbhar Sanchar: Self-Sufficient Communications Jay Jhaveri , Abhay Gupta , Prem Chhabria , Neeraj Ochani , Sharmila Sengupta , Mrs. Sunita Suralkar , and Shashi Dugad

Abstract In the light of recent war crimes and data piracy conspiracies, privacy is of utmost importance to an organization and even to an individual. The majority of the population is dependent on third-party services for their daily communication. Albeit these major corporations advertise “secure” means of chat transfer, they install various kinds of backdoors to sell the user’s data to advertisers. Under the notion of going “Aatmanirbhar” i.e., Make in India, we have developed an indie solution without incorporating any third-party services or APIs. “Aatmanirbhar Sanchar” aims at providing users with a real-time off-the-grid, secure, and anonymous messaging service. It features an End-to-End encrypted transmission of messages and data files likewise. This is achieved by combining the open-source AES algorithm with a self-developed XOR encryption process. Keywords Messaging · Security · Privacy · Self-sufficient · Aatmanirbhar · Self-hosting · AES · Hash-based message authentication code verification J. Jhaveri (B) · A. Gupta · P. Chhabria · N. Ochani · S. Sengupta · Mrs. S. Suralkar Computer Engineering, Vivekanand Education Society’s Institute of Technology, Mumbai, India e-mail: [email protected] A. Gupta e-mail: [email protected] P. Chhabria e-mail: [email protected] N. Ochani e-mail: [email protected] S. Sengupta e-mail: [email protected] Mrs. S. Suralkar e-mail: [email protected] S. Dugad Tata Institute of Fundamental Research (TIFR), Mumbai, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_41

545

546

J. Jhaveri et al.

(HMAC) · Indigenous private server · Scalability · Cross plat-form · Throwaway · Anonymous

1 Introduction In today’s world, privacy and security are of utmost importance to an individual. Let me elaborate: Data Privacy and Data Theft are the hot debate in the World-Wide mass media at the moment, but have you ever wondered what exactly it is? Have you ever questioned how the so-called “Free” applications are kept afloat? They pay their bills by selling that very data you unknowingly give them while using their “free” services. This borderline stolen data is then used for targeted, personalized advertisements and much worse. To combat this, we are developing anonymous communication software without the use of any third-party services, hence maximizing the privacy of an individual. To put it simply, secure messaging is a way of safely exchanging documents between users, healthcare providers, organizations, and their customers.

2 Motivation Currently, all other messaging services are hosted on Third-party cloud platforms, mainly Google and AWS cloud services. Let us consider WhatsApp for example, which has been recently acquired by The Facebook (META) team. After this acquisition, WhatsApp updated its privacy policy which gave access to Facebook to collect private information on its users causing many controversies and heated discussions in the IT industry around the globe. There is a dire need for us to focus on these privacy problems faced by users using these “free” applications like WhatsApp and Facebook messenger. The companies owning these applications do not take adequate security measures in handling the user data but drive their marketing/advertising agenda through the data provided by their users. Further, In the light of recent events, amidst the Russian-Ukraine War, there are major sanctions placed by the west on Russia, disabling them from using multiple western applications. Even their banking apps were restricted leading to a major downfall in their economy. Now, India has developed a native solution for the banking sector called the Unified Payments Interface, also known as UPI. Why not take this spirit and create an indie chat application? These were mainly our inspiration to create India’s very own messaging service without utilizing any kind of third-party services. In collaboration with the Tata Institute of Fundamental Research (TIFR), we have built a secure communication, cross-platform messaging application wherein a user can exchange vital information with other users and groups of users without being concerned about any kind of data leak or data monetization.

Aatmanirbhar Sanchar: Self-Sufficient Communications

547

3 Literature Review Cohn-Gordon et al. [1] in “A Formal Security Analysis of the Signal Messaging Protocol” explained that Signal Protocol is a private messaging protocol that provides instant messaging encryption to applications such as Skype, Facebook Messenger, and WhatsApp among many others, with more than 1 billion active users. The signal contains unique unfamiliar security features (such as “future privacy” or “postcompromise security”), which are made possible by ratcheting, a process through which session keys are updated with each new message. Singh et al. [2] in “Blockchain-Enabled End-to-End Encryption for Instant Messaging Applications” presented a blockchain-based E2EE framework for mitigating current messaging application vulnerabilities. During the installation of the application, the end-user device generates a pair of public/private keys and asks its mobile network operator to issue a digital Identifier and store it in the blockchain. The end user can obtain another user’s certificate from their chat private server and utilize a ratchet forward encryption process to interact securely with them. Botha1 et al. [3] in “A Comparison of Chat Applications in Terms of Security and Privacy, ECCWS 2019 18th European Conference on Cyber Warfare and Security” described a gadget that helps people adapt to social life by allowing them to understand domain messages, names, letters used in mailboxes, in daily newspapers, and so on. The major goal of the project is to solve the above problem by using a Raspberry Pi and an OCR sensor to recognize environmental messages automatically and then using TTS to translate those messages into voice or audio for better and easier engagement with society. Sabah et al. [4] in “Developing an End-to-End Secure Chat Application” presented a chat program that provides end-to-end security, allowing users to safely transmit confidential information without fear of data loss. In addition to the storage protection. This article presents a list of requirements for creating a secure chat application, and the program was created based on these requirements. The suggested chat application was compared to other popular apps based on those criteria, and it was also put to the test as a proof of concept for delivering End-to-End security. Burak [5] in “Encryption Methods and Comparison of Popular Chat Applications” proposed end-to-end encryption chat solutions that allow users to safely transmit personal information. The paper includes a list of requirements for developing a secure chat application. Canetti [6] in “Universally Composable End-to-End Secure Messaging” explained all the contemporary widely accepted encryption algorithms in detail and their limitations in the real practical world. It also helped in choosing the most suitable encryption algorithm for this chat application. Emura [7] in “Membership Privacy for Asynchronous Group Messaging” focuses on a method capable of hiding membership information from the viewpoint of non-group members in a secure group messaging (SGM) protocol, which we call “membership privacy”.

548

J. Jhaveri et al.

Fig. 1 WhatsApp Plain text Backup Proof

4 Lacuna in the Existing System “Data is the new fuel” and major tech corporations are utilizing every gizmo at their disposal to amass and utilize user data for monetary advantage because their customers’ personal and behavioral data is worth millions of dollars if mined to its full potential. WhatsApp and other Applications [8] provide the option of verifying users’ public keys, but the mechanisms used are not robust and pose major session hijacking issues. Besides this, there is no reliable third-party involvement to check the suitability of keys stored on WhatsApp servers [9–13]. The backup method utilized by WhatsApp does not provide real end-to-end encryption (see Fig. 1). The alternate copy is kept in plain text on the user’s cloud, depending on the user’s OS, such as iCloud, Google Drive, One Drive, and so on.

5 Methodology 5.1 Joining a Chat Room When a user visits the web application at http://aatmanirbhar-sanchar.live/, hosted on the private servers at Vivekanand Education Society’s Institute of Technology (VESIT), he is greeted with the homepage asking for a Username along with a Room key (RK) (see Fig. 2). The entered username will act as the main identity of that particular user for the ongoing session. The room key is the most significant aspect of the chat application. This key serves a dual purpose: 1. The hash of the key acts as the identity of a particular group chat created using the same.

Aatmanirbhar Sanchar: Self-Sufficient Communications

549

Fig. 2 Home page

2. The key itself is used in the encoding and decoding process of incoming and outgoing messages. When the user enters a room key (RK), the key first goes through an extensive algorithm to test the strength of the key. If the resulting strength is not up to the standards for secure communication, it will warn the user of a weak key and the user may decide if he wants to proceed or add a new key. A passphrase is highly recommended instead of a password to ensure utmost privacy while communicating. To keep it user-friendly, a random passphrase generator has been added. Now, when the user clicks the join button, the following processes occur (see Fig. 3):

Fig. 3 Joining process

550

J. Jhaveri et al.

1. The key is first hashed using the SHA-256 algorithm to get ready to be transmitted to the server. 2. NOTE: The plain text key never leaves the client side 3. The username, AES encrypted using the RK, is appended along with the hashed RK and is sent to the server. The user can now share this room key with the intended recipients to begin a secure private communication channel with him/her.

5.2 The Ephemeral Chat Room When a user joins a chat room using the shared RK, the active user counter is incremented and their encrypted username is broadcasted to all the users active in the room utilizing which a greeting message is displayed (see Fig. 4). Now every user is subscribed to the following events: 1. 2. 3. 4.

Join Response: Handles a new incoming user. Chat Response: Handles incoming text messages. File Response: Handles incoming files. Leave Response: Handles a user leaving.

When a user joins a chat room using the shared RK, the active user counter is incremented and encrypted.

Fig. 4 The chat room

Aatmanirbhar Sanchar: Self-Sufficient Communications

551

5.3 Encryption Process Aatma Sanchar uses a double-layered encryption process for achieving enhanced security. The first layer constitutes the self-developed XOR encryption process: This encryption system is based on the concept that if an object is XOR’ed by the same key twice it will revert to its original state. To make this viable in this innovative era of cybercriminals vs cybersecurity, multiple iterations of permutations and combinations on the original entity take place before further encoding. To put it simply, a text message is first converted to its binary format in the shape of matrices. These matrices are then shuffled and reshuffled to increase protection followed by undergoing the XOR process by the room key 10. This encryption system is based on the concept that if an object is XOR’ed by the same key twice it will revert to its original state (Fig. 5). To make this viable in this innovative era of cybercriminals vs cybersecurity we have added multiple iterations of permutations and combinations of the original entity. To put it simply, a text message is first converted to its binary format in the shape of matrices. These matrices are then shuffled and reshuffled to increase protection followed by it getting XOR’d by the room key [9]. This encryption layer is followed by the Advanced Encryption Standard (AES256) algorithm 11 to ensure privacy while maintaining efficiency. From the Graph in Fig. 6, it is clear that while there exist faster encryption algorithms other than AES, as the file size increases (which is a common situation in a messaging platform), AES easily comes out on top. Hence, AES-256 was selected as the second layer in this encryption process.

Fig. 5 Second layer encryption using the XOR matrix method

552

J. Jhaveri et al.

Fig. 6 Second layer encryption using the XOR matrix method

Finally, to ensure the integrity and authenticity of the transmitted message, the process of Hash-based Message Authentication Code verification (HMAC) is also practiced which guarantees tamper-proof messaging 12. This is done by creating a hash-based checksum using the combination of the room key and the data ready to be sent to the sender’s client. This checksum value is then recomputed on the receiver’s device, and if any discrepancy is detected it indicates that the message was tampered with (Fig. 13).

5.4 Transmitting a Message The users can either directly type or send a text message using the provided chat box, else a user can also attach files up to the 50 MB limit to be transmitted. 1.

Sending a text message (see Fig. 7): a. When a user types a message and hits the send button the message is encrypted using the Room Key (RK) utilizing the doubly layered encryption algorithm mentioned before. Further, the SHA-256 hashed RK and the encrypted Username are appended into a dictionary along with the encoded message. b. Next, we use the HMAC (hash-based message authentication code) algorithm to ensure the authenticity and integrity of the message being sent. The HMAC is generated using the above-created dictionary and the Room Key (RK).

Aatmanirbhar Sanchar: Self-Sufficient Communications

553

Fig. 7 Sending a text message

c. Finally, this encoded dictionary along with the HMAC appended to it is emitted through a socket to the server. 2.

Sending a File as an Attachment (see Fig. 8): a. A user also has the option to send any type of file as long as it is under the 50 MB limit. When he/she selects the file to be uploaded, the file is first converted into its binary (Base 64) format. b. This binary format is encrypted using the Dual layer encryption process and is stored in the dictionary along with its file name and file type. Finally, as in the text messaging process, an HMAC code is calculated utilizing the above

Fig. 8 File transfer process

554

J. Jhaveri et al.

dictionary and the room key. This is then sent using the same socket method as for a text message.

5.5 Receiving a Message After a user has sent the encrypted message to the server then broadcasts the message to all the users currently connected to that particular room. On the receiver’s end before any decryption process can start, the HMAC is recalculated on the client end utilizing the encrypted dictionary and if any disruption is found, an error is displayed in the chat box indicating the message was tampered with (Further explained in the Cryptanalysis part of the paper). After verifying the HMAC code, the actual decryption process starts: 1.

Receiving a text message (see Fig. 7): a. The incoming encoded dictionary is first decrypted using the AES algorithm followed by the reverse XOR method. b. This decrypted message is shown to the user in the chat box along with the decrypted username of the sender (see Fig. 4). 2. Receiving a file as an attachment (see Fig. 8): a. The incoming encoded dictionary is first decrypted using the AES algorithm followed by the reverse XOR method. b. This generates the file in its pure binary (Base 64) format. This binary file is then converted according to the MIME type into its original form. c. This original form is then converted into a blob link from where the receiver can download the same. If the File type is in a known multimedia format (Music, Image, Video), then the user is also given the option to preview the same within the chat box itself (see Fig. 9).

5.6 Leaving the Chat Room A user can press the “leave” button to securely exit the chat room. Once pressed, the client is unsubscribed from all the live-time events and finally emits an encoded dictionary constituting the hashed room key and the encrypted user name. This username is then broadcasted to every active user with a message indicating that this person left the chat room. Once left, the user cannot access the chat history and for enhanced security, the chats are never stored on the local device. If every user leaves a particular chat room, the session is destroyed in instantly.

Aatmanirbhar Sanchar: Self-Sufficient Communications

555

Fig. 9 File preview in the chat room (light mode)

5.7 Experimental Environment Details The following tools and technologies have been used in the development of Aatmanirbhar Sanchar: • Front-End Development: React JS HTML/CSS • Backend Development (server-side): Languages used: – NodeJS – Socket.IO Compatible Operating System: – Ubuntu (16.04) – Windows 10 Requirements: – A Public Static IP to host the messaging application. – Android Studio for mobile applications (Auto File Sync). – Maps and Google Sheets API

556

J. Jhaveri et al.

6 Applications 6.1 In Large Organizations as a Quick, Secure Chat Platform Our chat application being ephemeral does not store any of the chat data on the local client machine or the cloud server. This, apart from making our product more secure from any kind of unauthorized access to the local machines, also saves a lot of vital storage space in the cloud servers that can be essential for more important subjects. Being a throwaway chat application, it can be reused N number of times without any load on the server or the client.

6.2 Communications Where Security is of National Importance In matters where national security is of utmost importance, one cannot simply rely on the external, untrustworthy third-party application for secure communication [11]. Our chat application has up-to-date encryption algorithms along with our selfdeveloped XOR encryption process ensuring utmost protection without using any kind of third-party material. For instance, during a hostage situation, the officers in command can safely plan a rescue operation along with their subordinates without sacrificing the vital game plan to anyone intercepting the conversation.

6.3 Aatmanirbhar Samakraman: File Auto-Synchronization App Using our encryption process, we have also developed an auto synchronization file app useful in situations where the user repeatedly stores important readings in the format of a file in a selected directory and wants to securely upload the same to a remote server. The user first selects a specific directory to be continuously monitored by the app. He/she then selects another directory where he wants these files to be moved once uploaded to the server. He then presses the start syncing button to activate the background process (see Fig. 10). Now every time a new file is stored in that specific directory it triggers the application and the file are automatically uploaded to the remote servers and the locally stored file is moved to the second directory. This transfer of the file to the server takes place using the Multi-Part technology after being encrypted with our XOR-encryption process.

Aatmanirbhar Sanchar: Self-Sufficient Communications

557

Fig. 10 Aatmanirbhar Samakraman android application

Along with the file, the current location coordinates of the user are also sent which is then used to display a tracking history on a web-based map UI for easy analysis: (http://file.aatmanirbhar-sanchar.live/) (see Fig. 11).

7 Results In the end, we were successful in developing a secure, private, ephemeral chat application and deployed it on a private server hosted at our college, the VESIT campus. The web application is completely free from third-party services and is fully built upon open-source libraries. We also developed an encryption system from scratch. This XOR encryption process is nothing but dividing the binary data into matrices followed by shuffling and reshuffling of the data and finally the data being XOR’d by a predefined key (see Fig. 5). Finally, the output generated from the XOR method is passed to the Advanced Encryption Standard (AES-256) Algorithm in turn ensuring utmost security.

558

J. Jhaveri et al.

Fig. 11 Aatmanirbhar Samakraman live-map tracking

Finally, to ensure the integrity and authenticity of the transmitted message, the process of HMAC (hash-based message authentication code) verification is also practiced which guarantees tamper-proof messaging. To test the security of our product, we have tried the following Cryptanalysis techniques: 1. Snooping a. An intruder listens to traffic between two machines on a network in a snooping attack. We prevent this attack by only transmitting everything in an encrypted format. To an outsider, everything will look like gibberish (see Fig. 12). 2. Man in the middle attack (MITM) a. An attacker intercepts a message/key sent between two communicating parties through a secure channel in this sort of attack and tries to alter them. We can detect any tampering in the incoming messages due to the HMAC verification process. If any alterations are detected, a “decryption error” is shown to all the users indicating a tampered message (see Fig. 13). 3. Server attack a. In the case wherein an intruder has successfully gained access to the remote server would cause no harm to the privacy of the users. This is achieved due

Aatmanirbhar Sanchar: Self-Sufficient Communications

Fig. 12 Intercepted message as visible to an intruder

Fig. 13 Error is shown indicating message tamper

559

560

J. Jhaveri et al.

to the fact that no vital information is stored or even transmitted to the server without it being dually encrypted. So, the worse this intruder can do is shut down the server itself. 4. TCP-SYN flood attack a. TCP-SYN flood is a type of DDOS attack where the attacker starts pinging the server continuously from several different IP addresses by not providing Full information which is required by the server. Due to this, the server has to disconnect the running applications and wait for the partially opened connection started by the mugger, which can take enough resources to render the system unusable to authorized congestion.

7.1 Comparison of Results with Existing Systems One of the major differences between Aatmanirbhar Sanchar and other similar applications is the promise of keeping your data safe and keeping the application open source for anyone to verify its’ contents. The majority of the existing chat applications keep their source code proprietary and hide data mining loopholes in their terms and conditions (Fig. 14.).

Fig. 14 Data leaks in existing software

Aatmanirbhar Sanchar: Self-Sufficient Communications

561

Fig. 15 Decentralized chat application

8 Future Scope 8.1 Converting the Application to a Decentralized WEB3 Application (See Fig. 15) Now, instead of a centralized server, we can also theoretically use a blockchain network to convert the application into a web3 decentralized application. Blockchain is a type of database. In this database, data is stored in the form of blocks and these blocks are chained together to form a blockchain. When each block in the chain is added to the chain, it is given a precise timestamp. It is extremely difficult to modify the contents of a block after it has been put into the blockchain. This is because each block has its own hash in addition to the previous block’s hash. So, in this, if a new user installs the app for the first time a certificate will be generated which is then stored in a block and appended to the blockchain. This certificate will be used as a public key to the blockchain system. If a user wants to send a message to a particular person the certificate of the recipient is accessed from the blockchain. The message will be encrypted using the signal protocol before being sent. Now, user 1 verifies user 2’s certificate, and once verified, user 2 will be allowed to decrypt the message otherwise an error should be thrown.

562

J. Jhaveri et al.

9 Conclusion Secure and Private communication is a serious issue in today’s world. One is not able to communicate with their loved ones without being spied upon by the “Mark Zuckerbergs” of the world. There is a serious lack of an indie Secure Communication application that can be freely and securely used by public and private organizations. Hence, we are very excited to present to the world, “Aatmanirbhar Sanchar: An Ephemeral, Anonymous, Secure Chat Application” based on a custom-based encryption algorithm that can be easily hosted on one’s very own private servers without any external eyes watching over. Conflict of Interest Statement On behalf of all authors, the corresponding author states that there is no conflict of interest.

References 1. Cohn-Gordon K, Cremers C, Dowling B, Garratt L, Stebila D (2020) A formal security analysis of the signal messaging protocol. J Cryptol 33(4):1914–1983. https://doi.org/10.1007/s00145020-09360-1 2. Singh R, Tewari H (2021) Blockchain-enabled end-to-end encryption for instant messaging applications. https://arxiv.org/abs/2104.08494 3. Botha JG, Van ‘t Wout MC, Leenen L (2019) A comparison of chat applications in terms of security and privacy. In: 18th European conference on cyber warfare and security. University of Coimbra, Portugal 4. Sabah N, Kadhim JM, Dhannoon BN (2017) Developing an end-to-end secure chat application. Int J Comput Sci Netw Secur 5. Burak M (2021) Encryption methods and comparison of popular chat applications. Adv Artif Intell Res 52–59 6. Emura K, Kajita K, Nojima R, Ogawa K, Ohtake G (2022) Membership privacy for asynchronous group messaging. National Institute of Information and Communications Technology (NICT), Japan. 7. Canetti R, Jain P, Swanberg M, Varia M (2022) Membership privacy for asynchronous group messaging. National Institute of Information and Communications Technology (NICT), Japan 8. Marlinspike M, Perrin T (2016) The X3DH key agreement protocol. Signal 9. Whatsapp whitepaper (2021) WhatsApp encryption overview. Whatsapp. https://www.wha tsapp.com/security/WhatsApp-Security-Whitepaper.pdf 10. Perrin T, Marlinspike M (2016) The double ratchet algorithm. Signal 11. Daemen J, Rijmen V (1999) AES proposal: Rijndael. Rijndael Block Cipher 12. Bellare M, Canetti R, Krawczyk H (1996) Message authentication using hash functions—the HMAC construction. RSA Lab CryptoBytes 13. Hess A (2015) Encryption and cyber security for mobile electronic communication devices. Fed Bur Inv

A Meta Heuristics SMO-SA Hybrid Approach for Resource Provisioning in Cloud Computing Framework Archana and Narander Kumar

Abstract Cloud computing is an up-to-date model for distributing information processing utility and provides a large amount of resources through the internet. The major challenges affecting a cloud computing environment include resource provisioning and security. In this paper, we focused on resource provisioning mechanisms using Meta-heuristics techniques such as spider monkey optimization (SMO) and simulated annealing (SA). A simulated annealing algorithm helps to give a fine solution along with statistical promises for uncovering the best solution, yet it cannot notify whether the best solution is found. So it requires another method to overcome this drawback. This paper presents the Spider Monkey Optimization algorithm with Simulated Annealing (SMO-SA) to generate the best fitness value possible. The aim of the proposed hybrid algorithm is to generate the minimum fitness value by combining spider monkey optimization with simulated annealing to provision the resources dynamically. This paper also presents the step-by-step mathematical working of our proposed hybrid algorithm by applying it to the relevant data set and calculating the speedup factor as well as mean square error (MSE) value along with fitness value, which shows the effective impact of our proposed SMO-SA algorithm. Keywords Cloud environment · Spider monkey optimization · Simulated annealing · Resource provisioning · Meta-heuristics · Fitness value · Speedup factor · Mean square error

Archana · N. Kumar (B) Department of Computer Science, Babasaheb Bhimrao Ambedkar University (A Central University), Lucknow, UP, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_42

563

564

Archana and N. Kumar

1 Introduction Nowadays, the cloud environment is a progressive conception that provides greater utilization of numerous assets, and it involves cloud service providers to furnish the assets to a cloud customer [1–3]. So the major concern is to provide resources appropriately. Resource provisioning is a mechanism that plays a challenging task in the cloud computing environment [4, 5]. Here some new search nature-inspired meta-heuristics techniques are available, which can use to provide the resources appropriately.

1.1 Spider Monkey Optimization Swarm intelligence includes many optimization techniques, and its recently developed and popular technique is SMO, which establishes its place in the group of optimization techniques. SMO is categorized as a nature-inspired stochastic optimization method formed on the conception of fission–fusion social structure (FFSS). It is similar to other swarm intelligence techniques where each swarm updates its position by information sharing and continuous learning. SMO generates fitness value for each spider monkey to show how near the food source is. The global head is considered the finest solution of the entire group, and the local head represents each group’s best solution. It involves six stages: local head stage, global head, global head learning stage, local head learning stage, local head resolution stage, and global head resolution stage [6, 7].

1.2 Simulated Annealing An Annealing is a procedure in statistical mechanics where the system is gradually cooled to reach a position of low energy where they are very strong. Simulated annealing is based on the concept of annealing, and Kirkpatrick developed it in 1983. This algorithm has a greater impact on handling optimization problems using the annealing technique. We used an objective function in this method and attempted to minimize the solution. Here calculating the probability that the metal will leap to a larger energy level is given by S = e(−c/T ) c = change value of the objective function T = temperatur e. In the SA procedure, randomly generate the initial solution.

(1)

A Meta Heuristics SMO-SA Hybrid Approach for Resource …

565

The system’s temperature plays a similar effect as the temperature of the physical annealing process. SA may be at the starting phase, most worsening moves select but only improving ones are allowed [8]. By integrating the features of the SMO and SA approaches, this paper proposed a hybrid method (SMO-SA) to efficiently perform resource provisioning in the cloud environment.

2 Related Work A maximum VM placement with minimum power consumption (MVMP) technique is proposed to increase earned benefits through cloud server provisioning. This technique also attempts to reduce the power budget. Here simulated annealing approach is employed to resolve a bi-objective optimization problem [9]. A hybrid algorithm of the meta-heuristic approach like Ant Colony optimization and simulated annealing is proposed for dynamic provisioning resources. This proposed ACO-SA algorithm is presented in multitier applications to schedule jobs with reduced costs [10]. A three-layered model is proposed to minimize the consumers’ load power generation system based on the framework of the cloud environment [11]. A modified PSO algorithm is proposed to improve resource utilization. This proposed EPSO technique also maximizes the quality parameters and reduces the cost. This simulated annealing was incorporated with PSO to overcome the optimal local issue [12]. An artificial bee colony and ACO algorithm are used for dynamic resource provisioning. This is proposed to minimize the time optimization of provisioning and minimize the number of resources in multitier clouds. The result showed that ACO worked faster than other meta-heuristic algorithms [13]. A constrained continuous optimization problem solved by proposed modified version of SMO method named as constrained spider monkey optimization algorithm. This algorithm also contrasts with other constrained versions of the artificial bee colony, particle swarm optimization, and differential evolution [14]. A hybrid algorithm of meta-heuristic methods like particle swarm optimization and simulated annealing for provisioning the resources in the multitier applications. Here also showed the simulation results to compare the PSO-SA algorithm with the PSO and SA algorithm in multitier applications [15]. The experimental results showed better performance in comparison to other latest advanced algorithms. An Ageist Spider Monkey Optimization (ASMO) method is also proposed as a new form of SMO [16]. A detailed review analysis of SMO is presented to show the effective impact of SMO variants in real-world optimization problems [17]. Improved spider monkey optimization (ISMO) technique is proposed to improve the charge of convergence. This proposed approach is used in local leader stage to update the solution location [18].

566

Archana and N. Kumar

A modified spider monkey optimization method is proposed to refine the convergence rate. In this algorithm global search capability improved by using the metropolis basis from simulated annealing [19]. A SMO is proposed for optimized resource allocation and energy utilization, Green Cloud scheduling model is suggested. The proposed approach was tested on different parameters by performing cloud simulation. The simulated results showed the proposed approach’s effective energy consumption, response time and resource utility [20]. An ARPS (Autonomic resource provisioning and scheduling) framework is designed to encounter resource allocation problems. This framework is also joined with Spider money optimization to resolve a multi-optimization problem. Simulation results presented a better outcome of the proposed method in a cloud computing environment [21, 22]. A Three-layered model is proposed to minimize the cloud and fog framework’s generating system power and consumer load. Analysis proves that PSO-SA performs stronger than RR [23]. A hybrid approach with cat swarm optimization and simulated annealing (CSM-CSOSA) has been proposed to schedule the resources dynamically [24].

3 Proposed Hybrid Algorithm (SMO-SA) We suggested a hybrid method of SMO and Simulated Annealing (SMO-SA) for provisioning the resources appropriately. SMO use as a local and global seeking, and after finding gBest and lBest we employ SA to find all over gBest. Specifically, lBest and gBest have more chances to improve fitness value. The numerical working is given to state the working of the proposed SMO-SA algorithm. In this proposed algorithm, we used the following mathematical formulations to perform this Proposed SMO-SA, which are as follows.

3.1 Mathematical Formulation Population initialization popinit = rand(lower limit, upper limit)

(2)

By Eq. (2), here randomly initialize the population to perform the proposed SMOSA algorithm. popinit represents the initialization of the population. Location update in local head stage Perturbation rate. It requires updating the monkeys’ position in the local leader phase. If the randomly generated number is greater or equal to the selected perturbation rate, then update the position, otherwise, accept the previous one.

A Meta Heuristics SMO-SA Hybrid Approach for Resource …

567

per tur bation rate ∈ (0.1, 0.9)

(3)

SM newi j = S Mi j + rdno(0, 1) × (L Hk j − S Mi j ) + rdno( − 1, 1) × (S Mr j − S Mi j )

(4)

Here SM newi j represents the new position of the monkey. S Mi j shows the jth dimension of the spider monkey (SM), rdno represents the random number between the range, L Hk j shows the local leader position from the jth dimension of the kth group, S Mr j represents the randomly selected SM from the kth group. Function value. In this proposed algorithm following equation helps to calculate the function value. f unction value ( f v) = x12 + x22

(5)

Here x12 , x22 represents the dimensions of each monkey. Fitness value  fitv =

1 , 1+fvi

if fvi ≥ 0 1 + absfvi , if fvi < 0

 (6)

By using Eq. (6) calculate the fitness value of each monkey fitv shows the fitness value and fvi shows the function value of each monkey. Location update in global head stage Probability. Probability requires updating the position of the global head stage. If the randomly selected number is less than the probability, then update the SM position otherwise, consider the previous one.  pr obabilit y (P) = 0.9 ∗

f itvi max f it

 + 0.1

(7)

Here f itvi shows the fitness value of each monkey and max f it represents the better fitness of among all monkeys. Metropolis criterion c = f itvnew − f itv pr evious

(8)

 f itv SM

(9)

S = e(−c÷T )

(10)

T =

568

Archana and N. Kumar

Here S represents the Metropolis criterion. c shows the positive change in fitness value and f itvnew , f itv pr evious represent the calculated new fitness value and previous fitness value. T shows the temperature value, which minimizes at the end. f itv Shows the fitness value of all monkeys, and SM shows the total number of monkeys. If a new solution is not better than the previous one, then apply this criterion to decide whether to accept or reject the solution in this proposed algorithm. If the random generated number is less than S, then accept the point, otherwise refuse the point. Reduction factor and random solution. In this proposed SMO-SA algorithm randomly generate the random and reduction factor.

3.2 Algorithm of Proposed Hybrid Approach (SMO-SA) The flow chart of proposed hybrid (SMO-SA) algorithm is shown in Fig. 1. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23.

Initialize population and perturbation rate. Evaluate position, function value, and fitness value. Pick out global and local heads. Location update by local head stage.    S Mi j + rdno(0, 1) × L Hk j − S Mi j + rdno(−1, 1) × SM newi j = S Mr j − S Mi j Location update by global head stage SM newii = S Mi j + rdno(0, 1) × (G H j − S Mi j ) + rdno(−1, 1) × S Mr j − S Mi j Now applying SA on gBest and lBest value to search around the gBest and generate factor(c) and calculate the temperature freduction itv T = SM Generate a new random solution and calculate the objective fitness value If new solution < previous solution Then accept a new point Otherwise, apply the metropolis criterion (to accept or reject the current point) S = e(−c/T ) If random value fitvprevious , we have to check other conditions to accept or reject this point) c = f itvnew − f itv pr evious

(7)

Now applying the Metropolis criterion is needed to accept or reject the current point. Choose random no. in the range (0, 1) r = 0.83. Metropolis criterion, P[X2 ] = e(−c/T) = 1.0209. Here r < 1.0209, so we can accept the new point = (12, 9.6) and current temp is high. Iteration = 2. gBest = lBest = 0.0042, u1 = 0.92, u2 = 0.82. generate new design point: x1 = (−12, 20), x2 = ( −9, 29). r1 = −12 + 0.92 * (20 − (−12)) = 17.44, r2 = −9 + 0.82 * (29 + 9) = 22.16. Xnew = (17.44, 22.16). Now calculate fvx = 795.21, fitvx = 0.0012 (Here we can see fitvnew < fitvprevious ), c = fitvnew − fitvprevious = −0.0030 (Here f value is negative, so there is no need for a metropolis criterion, we can accept the current point and increase the iteration to j = 3, as given max j = 2) Stop the process.

5 Results and Discussion We mathematically analyzed the working of the proposed hybrid SMO-SA algorithm on the data set [25]. Here we can see that our proposed hybrid algorithm helps to minimize the fitness value and to give the best fitness value by giving more chances to improve their lBest and gBest value. This proposed hybrid algorithm gives better results for resource provisioning. Here we also calculate the speedup factor and MSE

574

Archana and N. Kumar

Table 7 Comparative analysis between SMO and proposed hybrid SMO-SA algorithm Algorithm

Fitness value

Speedup factor

MSE

SMO

0.0015

1.9969

0.01303

SMO-SA

0.0012

1.9974

0.01300

Fig. 2 Comparison between fitness value of SMO and proposed hybrid algorithm (SMO-SA)

value to show the better working impact of SMO-SA over SMO. Table 7 shows the comparative analysis between SMO and the proposed hybrid SMO-SA approach. Here 3 parameters have been calculated to compare the proposed approach with the existing one, i.e., fitness value, speedup factor, and mean square error (MSE). The calculated fitness value for the proposed SMO-SA algorithm is 0.0012, which optimizes the fitness value of SMO (0.0015). The speedup factor and MSE of the proposed SMO-SA algorithm are 1.9974 and 0.1300, respectively, while for SMO speedup factor is 1.9969, and MSE is 0.01303. The speedup factor of the proposed SMO-SA approach is maximized, and the MSE value is minimized, which shows the proposed approach’s better working than SMO. Figure 2 shows the graphical representation of the calculated fitness value of the proposed SMO-SA and also the existing one (SMO algorithm) as in Table 7. This figure analyzes the SMO-SA fitness value optimized compared to the SMO algorithm.

6 Conclusion We suggested working in cloud computing to solve resource provisioning issues. A new hybrid method SMO-SA is proposed for resource provisioning based on the combination of spider monkey optimization with simulated annealing. In this paper, we mathematically show that our proposed method gives a better result. This

A Meta Heuristics SMO-SA Hybrid Approach for Resource …

575

proposed method shows that provisioning resources based on the SMO-SA algorithm help minimize the fitness value, which leads to keeping down the processing flow and resources cost. In this paper, mathematical proving is given to state the effective working of the proposed SMO-SA approach. Here three performance parameters are considered, i.e., fitness value, speedup factor, and MSE. There may provide the simulation results on this proposed algorithm for resource provisioning and consider execution time, price, and memory allowance as resources for future work.

References 1. Kumar TS (2019) Efficient resource allocation and QoS enhancements of IoT with fog network. J ISMAC 02:101–110 2. Chandy A (2019) Smart resource usage prediction using cloud computing for massive data processing systems. J Inf Technol Digit World 2:108–118 3. Srivastava P, Khan R (2018) A review paper on cloud computing. Int J Adv Res Comput Sci Softw Eng 8:17–20 4. Kumar N, Kumar S (2019) Resource management to virtual machine using branch and bound technique in cloud computing environment. Soft computing: theories and applications. Advances in intelligent systems and computing, vol 742. Springer, pp 365–373 5. Singh S, Chana I (2016) Cloud resource provisioning: survey, status and future research directions. Knowl Inf Syst 49(3):1005–1069 6. Sumalatha K, Anbarasi MS (2019) A review on various optimization techniques of resource provisioning in cloud computing. Int J Electr Comput Eng (IJECE) 9:629–634 7. Sharma H, Hazrati G, Bansal JC (2019) Spider monkey optimization algorithm. Evolutionary and swarm intelligence algorithms. Studies in computational intelligence, vol 779. Springer, pp 43–59 8. Dubey K, Sharma SC, Aida A (2020) A simulated annealing based energy-efficient VM placement policy in cloud computing. In: International conference on emerging trends in information technology and engineering (ic-ETITE). IEEE, pp 1–5 9. Addya KS, Kumar A, Sahoo B, Sarkar BKS (2017) Simulated annealing based VM placement strategy to maximize the profit for cloud service providers. Eng Sci Technol Int J 20:1249–1259 10. Leninfreda A, Dhanyab D, Kavithac S, Ashwini M (2019) Hybrid algorithm for resource provisioning with low cost and time using improved cost-based algorithm in hybrid cloud computing. J Intell Fuzzy Syst 1–10 11. Yasmeen A, Javaid N, Rehman O, Iftikhar H, Malik MF, Muhammad JF (2018) Efficient resource provisioning for smart buildings utilizing fog and cloud based environment. IEEE, pp 811–816 12. Mani K, Krishnan RM (2017) Flexible cost based cloud resource provisioning using enhanced PSO. Int J Comput Intell Res 13(6):1441–1453 13. Eawna MH, Hamdy S, EI-Horbaty EM (2015) New trends of resource provisioning in multitier cloud computing. In: Seventh international conference on intelligent computing and information systems (ICICIS’15). IEEE, pp 224–230 14. Gupta K, Deep K, Bansal JC (2017) Spider monkey optimization algorithm for constrained optimization problems. Soft Comput 21:6933–6962 15. Eawna MH, Mohammed SH (2015) Hybrid algorithm for resource provisioning of multitier cloud computing. In: International conference on communication, management and information technology (ICCMIT). Elsevier, pp 682–690 16. Sharma A, Sharma A, Panigrahi BK (2016) Ageist Spider Monkey Optimization algorithm. Swarm Evol Comput 1–23

576

Archana and N. Kumar

17. Agarwal V, Rastogi R, Tiwari DC (2018) Spider Monkey Optimization: a survey. Int J Syst Assur Eng Manag 9:929–941 18. Swami V, Kumar S, Jain S (2018) An improved spider monkey optimization algorithm. Soft computing: theories and applications. Advances in intelligent systems and computing, vol 583. Springer, Singapore, pp 73–81 19. Hazratia G, Shannab H (2016) Modified spider monkey optimization. In: International workshop on computational intelligence (IWCI). IEEE, pp 209–214 20. Samriya JK, Kumar N (2022) Spider monkey optimization based energy-efficient resource allocation in cloud environment. Trends Sci 19(1):1–19 21. Kumar M, Kishor A, Abawajy J, Agarwal P (2022) ARPS: an autonomic resource provisioning and scheduling framework for cloud platforms. IEEE Trans Sustain Comput 7:386–399 22. Sharma Y, Taheri J (2020) Dynamic resource provisioning for sustainable cloud computing systems in the presence of correlated failures. IEEE Trans Sustain Comput (c) 1–13 23. Yashmeen A, Javaid N (2018) Resource provisioning for smart building utilizing fog and cloud based environment. In: 2018 14th international wireless communications & mobile computing conference (IWCMC). IEEE, pp 811–816 24. Gabi D, Ismail AB (2018) Hybrid Cat Swarm Optimization and simulated annealing for Dynamic task scheduling on cloud computing environment. J ICT 17(3):435–467 25. Kumar N, Kumar S (2018) Virtual machine placement using statistical mechanism in cloud computing environment. Int J Appl Evol Comput 9:23–31

A Comprehensive Study of Automation Using a WebApp Tool for Robot Framework N. Alok Chakravarthy and Usha Padma

Abstract The procedure of manual testing takes a lot of time. Automation is particularly desired since testing process is also error-prone due to its repeated structure. Robot Framework is a flexible tool that makes use of the keyword driven testing methodology. It is straightforward to use when high-level keywords may be created from current keywords. Although a command line interface makes it easy to integrate new test libraries, it is also possible to create customised test libraries in Python or Java using a straightforward library API. All these capabilities guarantee that Robot Framework can be used to execute test cases in a timely manner. This paper explains how a WebApp tool could be used to quickly automate testing procedures by reducing expenses and raising the overall functionality of the software. Results comparison of manual testing with automation testing shows that automated tests run on average 80.46% faster than manual tests. Keywords Software testing · Test cases · Test automation · Robot framework

1 Introduction The typical approach used by businesses is manual testing. However, this is gradually disappearing due to the rise of quality tools and web-based apps. Automated testing has recently grown to be an essential component of the software development life cycle. The software product’s functioning and all its components or use cases are tested using the test automation procedure. It offers an effective method for managing several product characteristics at once. There are several proprietary and open-source test automation resources available [1]. Many of the tools that are now present are best suited for unit tests carried either by programmers or quality analysts. It must also N. Alok Chakravarthy (B) · U. Padma R. V. College of Engineering, Bengaluru, India e-mail: [email protected] U. Padma e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_43

577

578

N. Alok Chakravarthy and U. Padma

offer several libraries for connecting to the back end of the device being tested, such as computer or mobile apps. Numerous testing tools offered by manufacturers are quite complex and employ open-source or closed-source programming language. Automating manual tests that already exist is comparable to a developer implementing a software in a coding language to automate any other manual operation [2].

2 Associated Works Automation of robotic processes and acceptability testing are done using the Robot Framework. It is a platform-neutral programme with a developing network of libraries and outside tools. Selenium is a free and open-source automation tool that is quite well known right now. Selenium has several advantages for test automation. By permitting recording and playback, web apps may simply be evaluated. They can also execute different scripts across different browsers. There is a lot of difference between both. Robot is an amazing test framework, making use of test libraries both standard and external for executing tests. On the other hand, Selenium is just a WebDriver that would generally need some support from test automation runners for the test execution. Robot framework is used for performing all kinds of automated tests, whether they are related to UI Testing or API level testing. Selenium module facilitates the Selenium framework support in case of Robot. Selenium is widely used for web app automation testing recently. Numerous programming languages, such as Java, Ruby, and Python, are supported by Selenium WebDriver. Both Selenium and Cucumber are powerful and comprehensive automation testing tools and are very useful for web application testing. But while Selenium is one of the best automation testing tools, using it can be a daunting task for a non-technical person. Therefore, more and more companies are using Selenium and Cucumber together for automation testing. There are several significant distinctions between cucumber and selenium. Cucumber is an automation tool for behaviour-driven programming, whereas selenium is an automation tool for web apps. Cucumber script development is also easier than selenium script generation. Selenium can function without cucumber also. For step-definition execution, cucumber relies on selenium. Robot automation framework is best suited for web. The python test automation framework is the ideal option for test automation projects. Robot Framework can test MongoDB, FTP, Android, and Appium. It includes the Selenium WebDriver library among other helpful test libraries. To assist it would be as expandable as possible, it offers a variety of APIs. Integration testing is done in software testing while the software and hardware parts are integrated to look for errors in the test. All through the test procedure, it is a continual process. For a quick release, it is recommended to do integration testing. Network management includes elements like fault management, which controls how network element alarms are handled, as well as troubleshooting, log analysis, and responding to fault circumstances. Performance

A Comprehensive Study of Automation Using a WebApp Tool for Robot …

579

management, which assesses data collecting, reporting, data processing, and overall performance monitoring, is another component. Also, security management is for user and policy access control, along with logs authentication and authorization [3]. File handling refers to the method used by a software to save the relevant information and data in a file. To reuse this information in any software, it may be obtained or retrieved from these files. There are occasions when a program’s outcome, after it has been compiled and executed, does not achieve the desired result. The output of the application is repeatedly examined in such circumstances. Now, it takes a lot of work for any coder to repeatedly compile and execute the same software. That is precisely why using file handling is beneficial. Few uses of file handling are reusability, saves time, commendable storage capacity, and portability. To expand these tests and create less mistakes, a tool that was both simple to use and powerful was required. The tool would need to work on any system. User testing was performed on Windows and Linux, while server testing was done on Linux and Solaris. The tool was completely platform independent. The primary emphasis was on regression testing of effective and complete tests. Even though most of the tests had already been run at least previously, running regression analysis proved problematic. When several servers were deployed upon which tests were conducted, the problem became even more difficult due to their diverse configurations [4]. The basic method was for all test cases: setup environment, begin system tracking, execute test suite, terminate platform traceability, verify code traces, and review system traces. It was critical not to overlook the creation of a review at the end with survey results, which could take a significant time and resources because it is important to modify the list of test cases, label those that failed, and make some notes on why they failed, which can take a bit of time for more test cases [5]. The initial plan was to create a simple shell script that would run all the tests and evaluate the findings from log files. As a framework, keyword-driven testing, which allows test scripts to be executed at a higher level of abstraction, was examined. The concept of keyword driven evaluation is comparable to that of a service or subroutine in coding, in which the same code may be performed with varying parameters [6], making it an ideal candidate for the needed automation. Figure 1 depicts its incredibly modular architecture design. After a thorough investigation, Robot Framework [7] was determined to suit all specifications. It was written in Python, which is available on all major resources. Therefore, all the conditions for 100% interoperability were satisfied. As indicated by the fact that it is referenced on [8], Robot Framework looks to be one of the rare free software tools that allows a multi-platform system and is consistently updated. Robot Framework is a multi-purpose framework that may be used with any software. The test results are given in a straightforward, customizable tabulated form. When initiated, Robot Framework analyses the findings, performs the test cases, and generates reports and outcomes. Test modules control the interaction with the topic under test. Modules can use programme interfaces directly or use drivers, which are subordinate test mechanisms [9]. What was required was a graphical user interface (GUI) that made generating and modifying test cases straightforward. Its primary aim is to provide a simple interface for the establishment and improvement of test

580

N. Alok Chakravarthy and U. Padma

Fig. 1 Architecture of robot framework

data in the Robot Framework [10]. Test cases are produced from HTML pages and use keywords from test suites to control the product under test, whereas test suites are created from files to make it simple to store into any version of system. It can introduce additional, more precise keywords by combining and arranging the existing ones. It is simple to automate web applications, which entails using numerous software automation testing techniques to gauge a website’s performance. A procedure called automated web application testing employs several software tools to rate a website’s functionality. The running of tests and comparison of actual test results with anticipated or projected outcomes are made possible using efficient automated testing tools for web application testing. Web application automation testing is also lacking on websites. Therefore, companies of today should employ test automation for web apps to guarantee that the websites function properly and provide perfect performance to consumers. Benefits of Web Application Automated Testing are test coverage improves, ensures quicker debugging, effective test results, increased testing speed, rapid feedback, early bug identification. The test data for the Robot Framework is tabular and can be formatted in HTML, TSV or Text. When the Robot Framework is launched, it chooses a test data file converter based on the file extension and processes the test file. HTML files are typically utilised. Four different forms of file structure are gathered as test data files, as shown in Table 1. Table 1 File structure of test data Test name

Used for

Settings

(1) Adding test libraries, resource files, and variable files to the system (2) Specifying meta information for test cases and test suites

Variables

Defining an all-purpose variable for the test data

Test cases

Making test cases out of the provided keywords

Keywords

Generating user keywords from lower-level keywords already in use

A Comprehensive Study of Automation Using a WebApp Tool for Robot …

581

3 Proposed Work 3.1 Test Suite Creation Making programs with an instantaneous return is one technique to reduce errors that occur when new tool utilization is initiated [11]. Developing scripts that won’t take too long to write but will undoubtedly reduce time spent on manual testing. More importantly, by writing the programs, users will have a deeper understanding of the tool’s operation and be able to come up with even better codes. Since these programs have previously provided some benefit, little is lost if they are destroyed. Robot Framework may be thought of as a script because it is built on keywords, and a keyword conjunction can create a new user keyword. Robot Framework comes with a few predefined modules, but as it is a Python-based tool, it is simple to add libraries created in Python or Java. All users must do is create their individual function and return a value. A feature of RIDE called keyword completion displays the keywords that are identified in the test suite, the source that is being modified, the imported files, or modules. Additionally, parameters are checked continuously for all recognised keywords [12]. One of the characteristics of the Robot Framework is detailed keywords. Since RIDE can construct keywords, it is sufficient to characterize the test case first before creating the keywords and populating them with actions. Another scenario is learning that a particular sequence must be utilised frequently. In that scenario, the sequence can be grouped and designated as a new keyword. RIDE makes it simple to do this; all that is required is to mark the series, at which point RIDE will extract the lines and, if necessary, automatically construct a new keyword. RIDE will update the series and modify the test case after establishing a new keyword. It is possible to preserve keywords and variable definitions in source files so that they may be utilised across many suites. If the keyword is applicable to more tests, moving it to a common resource is a smart idea [13]. In this approach, redundant labour is prevented, and those keywords may be utilised in future testing.

3.2 Test Case Execution Although it can run an entire test suite or just a few selected test cases from the RIDE GUI, it is essential to execute test cases from the command line so that their processing may be easily automated, for instance from a continuous integration server. This is often carried out in this manner as Robot Framework is a commandline programme. Different switches can be utilised in this manner. The specification of the key test cases is one of several things that may be stated. All important test cases must pass for the test suite to be successfully completed [14]. The background colour unquestionably indicates whether the entire test suite was successfully completed. Specification of important test cases must be done with discretion. Irrespective of the outcomes of other test cases, if important test cases pass correctly, the report

582

N. Alok Chakravarthy and U. Padma

Fig. 2 Test case report file

will be deemed acceptable. Statistics, on the other hand, will list and display the set of test cases that failed, if any. There is also a thorough report file created for additional manual inspection, which includes all actions, a thorough explanation of the processing parameters, and keyword output with activities that were incorrectly tagged. Since the term “Log” has been established, it is also feasible to add whatever is necessary to the log file. Since all results are already beautifully structured in HTML, using it for reporting is extremely handy. Additionally, the Robot Framework creates an XML output file that may be utilised for additional study [15]. After running the test suite, an HTML report is produced, as seen in Fig. 2. Figure 3 shows the test case log file which is generated as an output to analyse the results.

4 Result Analysis Every time, a test suite that is automated may examine the entire application. It will take a bit longer to go over everything manually. Therefore, when automation does discover an issue, it usually does so quickly after the wrong modification was applied. When there have just been a day’s number of modifications, testing is lot quicker, which also means inexpensive. Thus, the effectiveness of automation is increased. If automated tests are well-written, they may be executed in a certain

A Comprehensive Study of Automation Using a WebApp Tool for Robot …

583

Fig. 3 Test case log file

sequence that might change daily. This can be a quick and easy approach. Before Robot Framework, running the test suite required one candidate to run the test cases systematically, search for trails, and be occupied for couple of days [16]. With Robot Framework, the actual system only takes about an hour to complete, and since only one single operation is required, a user may focus on other things while the test suite is being executed. The new automated testing platform, which is the WebApp tool, will be beneficial and useful in all areas of testing. As technology advances, different new features may be implemented into the tool to provide further functionality. It is an essential component in developing a successful testing plan. Various advanced machine learning and artificial intelligence methods can be used to develop a feature that speeds up request execution and placement. The benefits of automating this execution procedure can be leveraged to compensate for the disadvantages of the human approach. In the future, the WebApp may be made more reactive to handle severe server demands, and alternative testing frameworks and methods can be introduced [17]. Because there is a chance for human mistake, the manual testing procedure is wildly inaccurate, but the automated process is trustworthy because it is script- and code-based. Additionally, automated testing is done quickly, but manual testing takes a lot of time. Table 2 depicts the statistics for the comparison between Manual and Automated testing. Results comparison of manual testing with automation testing using robot framework shows that automated tests run on average 80.46% faster than manual tests. Without the use of test scripts, a human executes the tests manually, step by step. Test automation frameworks, along with additional tools and software, are used to continuously perform tests during automated testing. The human tester performs it during

584 Table 2 Time comparison between manual and automated testing

N. Alok Chakravarthy and U. Padma Metrics

Manual (in hours)

Automatic (in hours)

Preparation of one test case

5:00

5:00

Execution of one test 0:02 case

0:02

Report for one test case

0:04

0:01

Total time used for one test case

5:06

5:03

For 100 test cases—one suite run

10:24 Command line interface

2:00 Machine interface

manual testing. The tool performs it during automated testing. Manual testing is difficult and slow. But it excels at handling complicated situations, which is a strength. Coding and test management are necessary for automated testing. On the bright side, it covers a lot more combinations and is considerably faster. Users often prefer automated testing compared to manual testing since it saves time. The time-consuming nature of manual testing led to the development of automated testing. Testing that is automated moves forward much more quickly.

5 Conclusion Working with the Robot Framework has the advantage of allowing users to write test cases that maintain an organic procedure, with prerequisites, execution, validation, and cleaning coming last. Real language is used for keyword definitions, making it easier for non-technical people to understand test cases. This, combined with its straightforward use and simple library extensions, make it an excellent tool for test case automation. Everything is reviewed periodically, and results are made and uploaded on the websites instantly as well. When the choice to incorporate continuous integration with a WebApp tool was taken, this also conserved a substantial amount of time. The easiest way to gauge the expense of automating a test is to look at how many manual tests it prevents from running and how many faults it creates [17]. This is undoubtedly the Robot Framework’s largest asset. Robot framework is easy to set up and use, and it facilitates in the implementation and development of test cases. Due to automating, manual work can be reduced while still generating error-free results. By reducing the requirement for manual testing, the expense of automating is reduced, which gives the robot framework its greatest strength. It is a very dependable and viable framework for writing and validating test scenarios. The produced report and log files aid in the rapid investigation and troubleshooting of issues. Manual testing

A Comprehensive Study of Automation Using a WebApp Tool for Robot …

585

can also be used to calculate the cost of this. Integration with the WebApp tool via the robot framework significantly minimizes human intervention time and effort. The drawbacks of automated testing are without a human element, it might be challenging to get understanding of the user interface’s visual elements, such as colours, fonts, and sizes. The price of the testing project may rise due to the potential expense of the automated testing technologies. Testing automation software is not yet error-free. Each automation technology has constraints that restrict the breadth of automation. Another significant challenge with automated testing is script debugging. It costs money to maintain tests. Some common challenges faced during automation include email assertions and asserting with background colours, as part of GUI testing, it is often necessary to visually verify if the colour has been modified after activities. While the background colour isn’t available at the tag level for certain web components, the style property and background colour are visible at the tag level for others. Additionally, file handling is required because the programme being tested includes capabilities like file upload and download. Values from external files must be read by the data-driven routines. A list of keywords for managing files in the operating system library is provided by the robot framework. The future scope involves new automated testing platform that will be beneficial and useful in all areas of testing. As technology advances, different new features may be implemented into the tool to provide further functionality. It is an essential component in developing a successful testing plan. Various advanced machine learning and artificial intelligence methods can be used to develop a feature that speeds up request execution and placement. The benefits of automating this execution procedure can be leveraged to compensate for the disadvantages of the human approach. In the future, the WebApp can be made more reactive to handle severe server demands, and alternative testing frameworks and methods can be introduced.

References 1. Lei B, Li X, Liu Z, Morisset C, Stolz V (2010) Robustness testing for software components. Sci Comput Progr 75(10):879–897 2. P. Laukkanen (2006) Data-driven and keyword-driven test automation frameworks. Master Thesis, Helsinki University of Technology 3. Mishra A, Jaiswal A, Chaudhari L, Bodade V (2022) Health record management system—a web-based application. J ISMAC 4 (2021):301–313 4. Rice RW (2003) Surviving the top ten challenges of software test automation. In: Proceedings of the software testing, analysis & review conference (STAR) East 2003. Software Quality Engineering 5. Joby PP (2020) Expedient information retrieval system for web pages using the natural language modeling. J Artif Intell 2(02):100–110 6. Lewis WE (2005) Software testing and continuous quality improvement. Auerbach Publications 7. Liu J-P, Liu J-J, Wang D-L (2018) Application analysis of automated testing framework based on robot. In: Third international conference on networking and distributed computing. IEEE 8. Yuste P, de Andrés D, Lemus L, Serrano JJ, Gil P (2017) Integrated Nexus-based real-time fault injection tool for systems. In: IEEE international conference on dependable systems and networks

586

N. Alok Chakravarthy and U. Padma

9. Batni NS, Shetty J (2020) A comprehensive study on automation using robot framework. Proc Int J Sci Res (IJSR) 9(7):1033–1036 10. Yadav V, Botchway RK, Senkerik R, Oplatkova ZK (2021) Robotic automation of software testing from a machine learning viewpoint. Soft Comput J 27(2):1–6 11. Yashaswini HR, Kulkarni P (2021) Robot automation framework for implementing and developing the requirements of network elements. Int J Adv Res Ideas Innov Technol 7(3):1–4 12. Poornachandra Tejasvi TM, Prakash KR (2019) Automation of NetAct integration using robot framework. IJERT Publications 13. Nagendra M, Chinnaswamy CN, Sreenivas TH (2018) Robot framework: a boon for automation. ISSN Publications 14. Hassan A, Hudec L (2018) Role based network security model: a forward step towards firewall management 15. Bhuvaneshwari, Hemanth Kumar AR (2017) Automation for configuration of mobile networks using robot framework 16. Monica, Shettar SN (2017) Survey on robot framework. IJERT Publications 17. Stresnjak S, Hocenski Z (2011) Usage of robot framework in automation of functional test regression. ICSEA Publications

Detection of Mirai and GAF-GYT Attack in Wireless Sensor Network Hanjabam Saratchandra Sharma, Moirangthem Marjit Singh, and Arindam Sarkar

Abstract Wireless Sensor Network plays an important role in collecting data from different environments where human involvement is deemed fatal or unnecessary. Apart from its usefulness, security threats and vulnerabilities exist as a common problem. A robust IDS (intrusion detection system) in WSN will be helpful in detecting and classifying the types of attacks, so as to remove or nullify the security threats. In this paper, we proposed a method to detect Mirai and GAF-GYT attacks in WSN using CNN along with f_classif function and normalization. Further, implementation of the proposed method has been carried out considering the scenarios: CNN without normalization along with f_classif function and CNN without normalization and without f_classif function. It is seen that the method that uses CNN along with f_classif function and normalization exhibits better performance in terms of parameters such as TPR, PPV, TNR, NPV, FPR, FDR, and FNR. Keywords Wireless sensor network · Convolution neural network · Mirai attack · GAF-GYT attack

1 Introduction A Wireless sensor network is a network formed by a collection of a huge number of a distributed autonomous sensor nodes, each node is a sensor device which consists of a sensor that measures various evironmental surrounding conditions such as heat, humidity, sound, pollution level, vibrations, etc., a transceiver, that receives or transmits the sensed or collected data to other neighboring sensor nodes, a battery that powers up the sensor device, a processor unit for computational purposes and a H. Saratchandra Sharma (B) · M. Marjit Singh Department of Computer Science and Engineering, North Eastern Regional Institute of Science & Technology, Nirjuli, Arunachal Pradesh, India e-mail: [email protected] A. Sarkar Department of Computer Science and Electronics, Ramakrishna Mission Vidyamandira, Howrah, West Bengal, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_44

587

588

H. Saratchandra Sharma et al.

storage unit that may be used to store limited data. WSNs are widely used in various areas such as health care monitoring, industrial monitoring, Environmental/Earth sensing, battlefields, and traffic controls. The nodes are resource-constrained and hence they have limited processing power, storage, and limited energy. WSN can experience various problems during deployment because of hardware, software, and environmental issues. The reliability of the WSN is in constraint at all times with various basic objectives including long term deployment, highly reliable data transmission [1]. There are many functions of the WSN and data analysis is one of the key functions of the WSN. And intrusion detection is one of the main objectives of data analysis [2].

2 Background There are several attacks that are encountered in the Wireless Sensor Networks. However, the paper focuses only on Mirai and GAF-GYT attacks which is briefly discussed in Sects. 2.1 and 2.2.

2.1 Mirai Attack Mirai is a malware that converts normally functioning networks into bots that are remotely controlled that can be manipulated to be part of a botnet in large-scale network attacks [3]. It attacks devices like home routers, cctv cameras, making them into attack network of remote controlled bot. Mainly used by cybercriminals to attack computer system in massive DDoS.

2.2 GAF-GYT Attack It is a malware that infect Linux System to launch DDoS. It is also known as Baslite. It attacks small routers in home and office to launch DDoS attacks. It targets various IoT and WSN and perform DDoS attacks at the same time. The GAFGYT mainly target the gaming industries [4].

3 Related Work Sarma [5] proposed a model based on DCNN to detect Mirai and GAF-GYT attacks in IoT. DCNN is applied for classification. The proposed model used MAIG for optimization of filters and its size in the convolution layer. The result of the experiment

Detection of Mirai and GAF-GYT Attack in Wireless Sensor Network

589

shows that the proposed model achieved higher accuracy in the detection of Mirai and GAF-GYT attacks as compare to methods such as DCNN, FAE-GWO-DBN and AIG. Liu et al. [6] formalized a multiple-mix-attack model. Then, PD (Perceptron Detection) was proposed that make use of perceptron and k-means to determine the trust values of the nodes of IoT and to detect malicious nodes. The route of the network is optimized and enhanced preceptron learning process named as PDE (Perceptron Detection with enhancement) is design to enhanced the detection accuracy. From the result it is shown that PD and PDE have higher rate of accuracy in detecting malicious nodes as compared to other similar methods. Baig et al. [7] have proposed an ADE-based attack detection scheme suitable for IoT. The paper presents a smart DoS detection strategy that consists of modules for generation of data, ranking the features, training, testing and generation. The proposed method was tested in real life condition using actual IoT attacks and is found to be superior from the other existing traditional classification techniques in terms of detection accuracy. Ismail et al. [8] proposed a multi-layer machine learning detection system to palliate different types of cyber-attacks that harms WSNs. Two models of machine learning detection system are deployed one at Base Station and another at the monitor nodes. For First-layer detection, Naive Bayes algorithm is used and for Secondlayer detection, LightGBM algorithm is used for binary classification and multi-class classification respectively. Internal DoS (four network-layer) attacks are detected using the proposed system. Umamaheshwari et al. [9] proposed a decision tree based attack detection system. To reduce the detection time, they experimented with different feature selection algorithms such as Correlation Score, Fisher Score, Relief and Minimum Redundancy maximum relevance (MRMR) algorithms. And from the results, MRMR feature selection performed superiorly from the other algorithms and yields 98.58% in accuracy, 92.81% in sensitivity, 98.46% in specificity, 93.86% in precision and 15.12 s training time. Nguyen et al. [10] proposed a IOT botnet detection method based on PSI-rooted subgraph based feature. Their method shows superior performance as compared to Decision Tree, Bagging, KNN, Random Forest and SVM. Jung et al. [11] proposed a CNN-based deep learning detection system that detect IoT botnets that are malicious. Rathore et al. [12] proposed an attack detection system based on fog computing. NSL-KDD dataset is used in their method. Their method shows better detection accuracy as compared to centralized attack detection process. Singh et al. [13] proposed a technique that detect wormhole attacks in wireless sensor network using artificial neural network.

590

H. Saratchandra Sharma et al.

4 Proposed Method In the proposed method, we used the dataset available from the UCI repository [14] to detect Mirai and GAF-GYT attacks in WSN. The proposed method consist of CNN for classifying the attack and f_classif function for selecting the best features from the dataset. After data preprocessing the normalization of data is carried out in order to: remove data redundancy and set a similar value of range of the features. After normalization phase, f_classif function is applied to the normalized dataset to select the best 20 features from the dataset. Thereafter, attack classification is performed using CNN and finally, performance is undertaken using the parameters such as TPR, PPV, TNR, NPV, FPR, FNR, and FDR. The flowchart of the proposed method is shown in Fig. 1.

4.1 Attack Detection System with F_classif Feature Selection The principal motive of the proposed research work is based on the detection of different types of attack in WSN, here, an attack detection method is formed using two stages: selection of features using f_classif and Classification. The dataset that we are working on is obtained from the dataset [14] that have 9 applications. From among the 9 applications we worked on 7 applications namely Provision_PT_737E_Security_Camera, Ecobee_Thermostat, SimpleHome_XCS7_1002_WHT_Security_Camera, Samsung_SNH_1011_N_Webcam, Fig. 1 Flowchart of the proposed method for detection of Mirai and GAF-GYT attack

Detection of Mirai and GAF-GYT Attack in Wireless Sensor Network

591

Phillips_B120 N10_Baby_Monitor, Ennio_doorbell, Danmini_Doorbell and SimpleHome_X CS7_1003_WHT_Security _Camera. These datasets are prepared for pre-processing, here normalization is performed, that convert the data in the range of 0–1. The process of normalization is discussed in brief in the following.

4.2 Normalization The process is mainly performed before the feature extraction operation. It is a scaling method that is mainly implemented in the data preparation stage to convert the values of numeric columns to a common scale in the dataset. The Normalization is explained in Eq. 1. 

N=

ei j   max ei j

i = 1 to M,



(1)

j = 1 to N

After the normalization, features with best scores are selected using the f_classif method from the normalized data Z = {z1 , z2 , …, zn }. And classification process is carried out on the selected or extracted features using CNN.

4.3 Feature Extraction Using f_classif Function Features are selected using the f_classif function. The f_classif() is mainly utilized in extracting the best features (that have the highest value or score) through the SelectKBest class. ˙It is a method which is obtainable from the scikit-learn that uses a scoring function and and based on the scores of the features, the features are positioned accordingly [15].

5 Results and Discussion 5.1 Experimental Setup The proposed detection technique was implemented in Jupyter Lab using scikitlearn. Seven applications were used that was retrieved from dataset given in [14]. For convinience the proposed method is abbrebriated as N and F_classif and the other two scenarios that is NC (method that uses CNN without normalization and without f_classif function) and FA method that uses CNN without normalization

592

H. Saratchandra Sharma et al.

along with f_classif function). Here, two analyis was done under GAF-GYT attack and Mirai attack. The performance of the method N and F_classif was compared with a method NC and a method FA. Positive and negative measures were taken into account while carrying out the analysis. Positive measures includes sensitivity or TPR (True Positive Rate), specificity or TNR (True Negative Rate), precision or PPV (Positive Predictive Value), and NPV (Negative Predictive Value). And negative measures that includes FPR (False Positive Rate), FDR (False Discovery Rate), and FNR (False Negative Rate).

5.2 Performance Analysis with Positive Measure Under GAF-GYT Attack The performance of the work is compared for 7 applications regarding the positive measures for the detection of the GAF-GYT attack (Fig. 2). The performance is found to be superior, if it maintained a higher score when juxtaposed to other models. For sensitivity or TPR measurement, the method N and F_classif achieves better performance over FA and NC by 0.81 and 33.21%, respectively. For specificity or TNR measurement, the method N and F_classif achieves better performance over FA and NC by 0.0048 and 10.64%, respectively. The method N and F_classif achieves better performance over FA and NC by 0.0097 and 6.375%, respectively for precision or PPV. And finally for NPV, the method N and F_classif achieves better performance over FA and NC by 0.93 and 16.58%, respectively. Thus, method (N and F_classif) performed superiorly over FA and NC for all positive measures.

5.3 Performance Analysis with Negative Measure Under GAF-GYT Attack The method’s performance is compared for 7 applications regarding the negative measures for the detection of the GAF-GYT attack (Fig. 3). The performance of a system under consideration is found to be better, if the result observed of negative measures is minimum. While considering the FPR measure the method N and F_classif is found to be −0.0035 and −10.64% lesser than FA and NC methods. For FNR measure the method (N and F_classif) is found to be −0.81 and −33.21% lower than that of FA and NC methods. And finally for FDR, the method N and F_classif achieves lower performance over FA and with NC by −0.01 and −6.37% respectively. Thus, from the above observations, the performance analysis of the method N and F_classif with respect to negative measures under GAF-GYT attack is low. Hence, we can deduced from the above observation that method (N and F_classif) performed much better that the other two methods.

Detection of Mirai and GAF-GYT Attack in Wireless Sensor Network TNR

TPR

1

1

0.5

0.5

0

0

NC

FA

593

NC

N and F_classif

(a)

FA

N and F_classif

(b) NPV

PPV

1

1

0.5

0.5

0

0

NC

FA

NC

N and F_classif

(c)

FA

N and F_classif

(d)

Fig. 2 Models’ performance under GAF-GYT attack for positive measures a TPR b TNR c PPV d NPV FPR

FNR

1

1

0.5

0.5

0

0

NC

FA

N and F_classif

(a)

NC

FA

N and F_classif

(b) FDR

0.3 0.2 0.1 0

NC

FA

N and F_classif

(c) Fig. 3 Models’ performance under GAF-GYT attack for negative measures a FPR b FNR and c FDR

594

H. Saratchandra Sharma et al. TPR

TNR

1

1

0.5

0.5

0

0

(a)

NC

FA

N and F_class

(b)

NC

PPV

1

0.5

0.5

0

0

NC

FA

N and F_class

N and F_class

NPV

1

(c)

FA

(d)

NC

FA

N and F_class

Fig. 4 Models’ performance under Mirai attack for positive measures a TPR b TNR c PPV and d NPV

5.4 Performance Analysis with Positive Measure Under Mirai Attack For sensitivity or TPR measurement, the method N and F_classif achieves better performance over FA and with NC by 2.25 and 22.65%, respectively. For specificity or TNR measurement, the methodN and F_classif achieves better performance over FA and NC by 1.14% and 28.31%, respectively. The method N and F_classif achieves better performance over FA and NC by 1.17% and 14.05%, respectively for precision or PPV. And finally for NPV, the method N and F_classif achieves better performance over FA and NC by 1.17% and 9.04%, respectively. Thus, method N and F_classif performed superiorly over FA and NC for all positive measures in the detection of Mirai attack (given in Fig. 4).

5.5 Performance Analysis with Negative Measure Under Mirai Attack On observing the FPR measure the method N and F_classif is found to be −1.14% and −28.31% lesser than FA and NC methods. For FNR measure the method (N and F_classif) is found to be −2.25% and −22.65% lower than that of FA and NC

Detection of Mirai and GAF-GYT Attack in Wireless Sensor Network FPR

FNR

1

1

0.5

0.5

0

0

(a)

NC

FA

N and F_class

595

(b)

NC

FA

N and F_class

FDR

0.4 0.3 0.2 0.1 0

(c)

NC

FA

N and F_class

Fig. 5 Models’ performance under Mirai attack for negative measures a FPR b FNR and c FDR

methods. And finally for FDR, the method N and F_classif achieves lower performance over FA and with NC by −1.17% and −14.05% respectively. Thus, from the above observations, the performance analysis of the method N and F_classif with respect to negative measures under Mirai attack is low. Hence, we can conclude that method N and F_classif performed better in detection of Mirai attacks as compare to the other methods (given in Fig. 5). In the approached method N and F_classif we use CNN along with f_classif function to select the best features from the dataset. ˙In other scenarios FA we use CNN without normalization along with f_classif function and in NC method, we simply use CNN without normalizing the dataset and without f_classif function. And from the above analysis we can conclude that method N and F_classif performed much better as compare to the FA and NC.

6 Conclusion An approach to detect Mirai and GAF-GYT attacks in WSN has been proposed using CNN and f_classif function. The attack detection technique is formed using two stages: feature selection using f_classif and classification. From among the 9 applications we worked on 7 applications namely Provision_PT_737E_ Security_Camera, Ecobee_Thermostat, SimpleHome_XCS7_1002_WHT_ Security_Camera, Samsung_SNH_1011_N_Webcam,

596

H. Saratchandra Sharma et al.

Phillips_B120N10_Baby_ Monitor, Ennio_doorbell, Danmini_Doorbell and, SimpleHome_XCS7_1003 _WHT_Security_Camera. Further, implemetation of the proposed method has been carried out considering the scenarios: CNN without normalization along with f_classif function and CNN wihout normalization and without f_classif function. From the results it is seen that method N and F_classif exhibits better performance as compare to NA and NC in the detection of Mirai and GAF-GYT attacks. For TPR measurement, the method N and F_classif achieves better performance over FA and NC by 0.81% and 33.21%, respectively, for TNR, by 0.0048% and 10.64%, 0.0097% and 6.375%, respectively for PPV. And for NPV, by 0.93% and 16.58%, for FPR the method N and F_classif is found to be −0.0035% and −10.64% lower than FA and NC methods, −0.81% and −33.21% lower than FA and NC methods for FNR. And finally for FDR, −0.01 and −6.37% lesser than FA and NC respectively under GAF-GYT attack. ˙In case of Mirai attack detection, TPR for N and F_classif is greater than FA and NC by 2.25% and 22.65%, respectively, for TNR, 1.14% and 28.31%, respectively, 1.17% and 14.05%, respectively for PPV. And for NPV, the method N and F_classif achieves better performance over FA and NC by 1.17% and 9.04%, respectively, for FPR measure the method N and F_classif is found to be −1.14% and −28.31% lesser than FA and NC methods, −2.25% and −22.65% lower for FNR. And for FDR, the method N and F_classif achieves lower performance over FA and with NC by −1.17% and −14.05%, respectively. Although, it is noted that the performance of the approached method N and F_classif is better in comparison to the other two scenarios, the performance of the proposed method can be enhanced by using an optimizer to optimized filter size and filter count in the convolution layer and as well as the activation function selection. The selection of the use of perfect otimizer is kept for future work.

References 1. Jurdak R, Wang XR, Obst O, Valencia P (2011) Wireless sensor network anomalies: diagnosis and detection strategies. Intell-Based Syst Eng 309–325. https://doi.org/10.1007/978-3-64217931-0_12 2. OReilly C, Gluhak A, Imran MA, Rajasegarar S (2014) Anomaly detection in wireless sensor networks in a non-stationary environment. IEEE Commun Surv Tutor 16(3):1413–1432. https:// doi.org/10.1109/surv.2013.112813.00168 3. Das S, Amritha PP, Praveen K (2021) Detection and prevention of mirai attack. In: Reddy VS, Prasad VK, Wang J, Reddy KTV (eds) Soft computing and signal processing. Advances in ıntelligent systems and computing, vol 1325. Springer, Singapore. https://doi.org/10.1007/ 978-981-33-6912-2_8 4. Hu X, Sun R, Xu K, Zhang Y, Chang P (2020) Exploit ınternal structural ınformation for IoT malware detection based on hierarchical transformer model. In: 2020 IEEE 19th ınternational conference on trust, security and privacy in computing and communications (TrustCom), 2020, pp 927–934. https://doi.org/10.1109/TrustCom50675.2020.00124 5. Sarma SK (2021) Optimally configured deep convolutional neural network for attack detection in ınternet of things: ımpact of algorithm of the ınnovative gunner. Wirel Pers Commun 118(1):239–260. https://doi.org/10.1007/s11277-020-08011-9

Detection of Mirai and GAF-GYT Attack in Wireless Sensor Network

597

6. Liu L, Ma Z, Meng W (2019) Detection of multiple-mix-attack malicious nodes using perceptron-based trust in IoT networks. Futur Gener Comput Syst. https://doi.org/10.1016/ j.future.2019.07.021 7. Baig ZA, Sanguanpong S, Naeem Firdous S, Nhan Vo V, So-In C (2020) Averaged dependence estimators for DoS attack detection in IoT networks. Futur Gener Comput Syst 102:198–209 8. Ismail S, Dawoud D, Reza H (2022) A lightweight multilayer machine learning detection system for cyber-attacks in WSN. In: 2022 IEEE 12th annual computing and communication workshop and conference (CCWC), 2022, pp 0481–0486. https://doi.org/10.1109/CCWC54 503.2022.9720891 9. Umamaheshwari S, Kumar SA, Sasikala S (2021) Towards building robust ıntrusion detection system in wireless sensor networks using machine learning and feature selection. In: 2021 international conference on advancements in electrical, electronics, communication, computing and automation (ICAECA), pp 1–6. https://doi.org/10.1109/ICAECA52838.2021.9675609 10. Nguyen H-T, Ngo Q-D, Nguyen D-H, Le V-H (2020) PSI-rooted subgraph: a novel feature for IoT botnet detection using classifier algorithms. ICT Express (in press, corrected proof, Available online 7) 11. Jung W, Zhao H, Sun M, Zhou G (2020) IoT botnet detection via power consumption modelling. Smart Health 15:100103 12. Shailendra Rathore J, Park H (2018) Semi-supervised learning based distributed attack detection framework for IoT. Appl Soft Comput 72:79–89 13. Singh MM, Dutta N, Singh TR, Nandi U (2020) A technique to detect wormhole attack in wireless sensor network using artificial neural network. In: Suma V et al (eds) Evolutionary computing and mobile sustainable networks. Lecture notes on data engineering and communications technologies, vol 53. Springer, Singapore, pp 297–307. https://doi.org/10.1007/978981-15-5258-8_29 14. https://archive.ics.uci.edu/ml/datasets/detection_of_IoT_botnet_attacks_N_BaIoT 15. Pathan MS, Nag A, Pathan MM, Dev S (2022) Analyzing the impact of feature selection on the accuracy of heart disease prediction. Healthc Anal 2:100060. https://doi.org/10.1016/j.hea lth.2022.100060,ISSN2772-4425

A Brief Review of Network Forensics Process Models and a Proposed Systematic Model for Investigation Merly Thomas and Bandu Meshram

Abstract Network forensics is a branch of Digital Forensics concerned with analysing the network traffic to see if any anomalies are present that may indicate an attack or could lead to one. The goal is to figure out what kind of attack it is by capturing the details, store them in a forensically sound manner, analyse, and then present them in some visual form. A model based on traceability and scenarios, with proven literature and justification is desired. This study offers a professional digital framework in which the investigative process model enhances the systematic tracking of offenders. Cyber fraud and digital crimes are on the rise, and unfortunately less than two per cent is the conviction rate worldwide. Continuous and scientific research in this area is crucial to ensure safe and secure internet usage especially for money transfers and confidential personal communication. This paper examines the essential development phases of a Network forensics investigation model, and compares different network and digital forensic methods, and also offers a systematic model of a digital forensic model for cybercrime investigation. The survey also includes classifications based on infiltration detection systems, trace backs, distribution models, and attack maps. The aim of this study is to facilitate the digital forensic process and identify improvised practices. The Systematic Network Forensic Investigation model (SNFIM) aims to establish appropriate policies and procedures for practitioners and organizations. Keywords Network forensics · Digital forensics · Infiltration detection systems · Traceback · Distribution · Attack maps

M. Thomas (B) · B. Meshram Veermata Jijabai Technological Institute, Mumbai, India e-mail: [email protected] B. Meshram e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_45

599

600

M. Thomas and B. Meshram

1 Introduction In today’s world of technology, advances in computers, small communicating devices, and networks have made our society increasingly dependent on cuttingedge information and communication technology and mobile appliances for personal activities, business transactions, and government services [1]. Identity theft, cyberbullying, data leakage or information denial, malicious software destroying the normal functions of the systems, distributed denial of service (DDoS) orchestrated through botnets, and malware targeting specific appliances such as smart vehicles, are all examples of new threats and cybersecurity issues widespread in the networked world. As a result of such incidents, the perpetrators are more intelligently and strategically identified, which makes it easier to perpetrate crimes and pursue illegal intentions [2]. Internet crimes are rampant, complicated and worth all the attention of academicians and researchers. Annual statistics paint a bleak picture of the number of people affected by cybercrime and the rate at which it is growing. It’s no wonder that cyber-crime is on the rise, and as a result, many people worldwide are researching this field of Forensics, and enhanced methods has to constantly evolve to keep pace with the strategic moves and advances of the perpetrators [3]. When the nodes are decentralized a decentralized solution may be a more efficient solution in grey and black hole attack settings by solving the problem of message overhead by appropriate measures [4]. Digital forensic can be put into two broad categories as follows [5]: • Computer forensics • Network forensics. As the Internet grows and the global spread of net-enabled devices increases, computer-related formal surveys are needed to get a grip of the situation. As a part of crime research, surveys describe the characteristics of a large population, map the entire networks in question, and acquire the knowledge necessary to administer such an extensive infrastructure. Crimes such as computer policy violations, fraud, email, and social media harassment, cyber-stalking, privacy, and terrorism need to be recorded on devices. Law enforcement bodies, network administrators, and criminal investigators rely on computer-assisted skills to investigate criminal and civil cases. Generally, digital forensics examines data that can be collected and analysed from device forensics or other residues, like when an archaeologist digs a site [6]. Information regarding a crime may already be recorded and available on a disk or device, but finding its location and accessing it is not easy. In its place of computer surveys, network expertise provides in sequence about how a user, or hacker gains the right of entry to the network. When a user or attacker logs in, criminal investigators use log files to explore, in sequence, what steps they took to commit the crime [7]. Finding the different URLs they visit and the login credentials used by a user or attacker may be found by different tools and multiple experts. Digital forensic analysts work as a team with experts from various fields to secure computers and network assets for an enterprise [8]. Computer literacy is one of the

A Brief Review of Network Forensics Process Models and a Proposed …

601

Fig. 1 Forensic investigation triangle Assessment

Investigations

Intrusion response

three triangular boundaries that ensure corporative computer defence and investigation. The triangle given in Fig. 1, the three sides depict the combined efforts of the team closely monitoring the assets of an organisation [9]: • Vulnerability assessment and risk management • Network intrusion detection and incident response • Computer investigation. Susceptibility appraisal and risk management experts have network monitoring, interruption discovery, and a team of digital investigators investigating forensic evidence in a suspicious context in order to gather digital confirmation connected to an occurrence or offence. Network expertise can be alienated into two distinct streams [10]. Network attacks can be divided into the following types: • Probing: (Surveillance and Interception)—Probing is a type of attack in which an attacker explores a network for vulnerable information or to find known flaws. An attacker with a map of available services on a network can use this data to seek exploits. Examples for this type of attack include Mscan, Nmap, Pingsweep and Port sweep attacks. • DOS: Denial of service (Interruption): An attack which involves flooding where attackers make computing or memory resource of the target machine too busy or too full to handle legitimate requests, thereby denying legitimate users access to the machine, e.g., syn flood. • User to Root Attack (U2R): An exploit in which the attacker who has access as a normal user account of the system exploits some vulnerabilities to gain root or superuser access privileges of the system, e.g., Buffer overflow attacks. • Remote to Local Attack (R2L): An attack instance where an attacker without having a user account gains access to a remote machine and sends packets. To get access as a user of that machine, the attacker exploits a vulnerability, such as guessing the password. During the forensic analysis of any network, a team of investigators prepares and investigates the incidence, to obtain information on a public or confidential network and find the source of these attacks. In addition, the data collected by the Network Inquiry Committee, a number of facts and digital evidence are reconstructed that can be submitted to law enforcement agency throughout production [11]. This may

602

M. Thomas and B. Meshram

include searching of digital devices to generate evidence related to an event that occurred during an investigation. As the quantity of digital data increases, analyzing data becomes a time-consuming process that may affect the legitimacy of the whole process. Existing search tools offer numerous search strategies for document analysis, but no document of interest addressing the magnitude of the types and number of cyber-attacks, could be found [12]. Thus, in the present scenario when a digital forensic tool is introduced into the system, investigators obtain a set of entry permit that are meaningfully related to the subject of the investigator’s interest [13]. 1. Identification phase: determine the data, mechanism and various entities related to the crime. 2. Image Acquisition and Preservation: preserving the crime scene by protecting it from any damage. 3. Data Recovery: collecting the relevant data or information interrelated to the crime. 4. Analysis: The researcher will perform a comprehensive analysis on the data by the theatre pattern matching; filter techniques by utilising a mixture of forensic tools. 5. Reporting: The canvassers summarize and present his hypothesis [14] (Fig. 2). In recent years, with computer crime, even in ordinary crime, important evidence or information has been increasingly stored in various electronic media, such as computers or smartphones. Copying digital data can be achieved by a few key strokes and almost impossible to distinguish from the original. Additionally, electronic data can be easily created, modified, or deleted from the original. Therefore, criminal investigations require high-level as well as diverse technologies to obtain sustainable evidence from digital data on suspects’ computers or smartphones [15]. Fig. 2 Forensic process model

A Brief Review of Network Forensics Process Models and a Proposed …

603

2 Related Works Sathwara et al. [16] have studied a forensic mechanism to investigate IoT-related crimes. IoT poses a number of challenges for forensic analysts. For example, in differentiation between private networks through public networks, the blurred lines between networks are increasingly disappearing. In addition, the testing of IoT tools is further complicated by the integration of diverse tools of IoT forensics, as well as the compatibility of identified and stored devices. The reason to include this paper is the supply of an IoT expert structure. Their goal was to support the digital exploration of IoT devices and to overcome the growing challenges of digital forensics and connection development. Shrivastava et al. [17] have studied Digital forensics is the systematic retrieval of evidence collected as a consequence of the investigation of critical data. After examining all the historical approaches used in the current models, their advantages and disadvantages are presented and an approach to the Forensic Model (EAF) is suggested; It connects all stages of the digital quest in detail. Valjarevic et al. [18] shows how to implement a prototype to meet all needs. In their previous works, the authors have suggested a comprehensive and relevant digital forensic procedure, which is used as the basis of the prototype. There is a sample software with two main functions. The first function is the role of an expert organization that can be used to lead and train new operators. The second function is to reliably document all the steps engaged within the specified process in a complete and relevant model of the digital forensic process. The merits of the investigation model include better improvements in the efficiency and effectiveness of the investigation. Omeleze et al. [19] have studied The Harmonised Digital Forensic Investigation (HDFI) progression representation is at present being developed as an international standard for digital forensics (ISO/IEC 27043), so it needs to be veteran. In this paper (HDFI) the process model is tested with an Android changeable cell phone. Kebande et al. [20] have introduced cost-effectiveness in companies by increasing the availability of digital devices in their day-to-day business operations. However, the development of this technology threatens many experts, because in the absence of an effective security model, digital forensic potential can be used to plan and prepare security events before an organization approved by the BYOD. Burrows et al. [21] investigates forensic investigative tools used in various steganography techniques have been developed for steganography samples and steganography detection and extraction for smartphones/mobile devices. Charles et al. [22] examine the internet forensic examination process available at the university and determine its ability to conduct a comprehensive digital forensic examination. Kao et al. [23] have shown that the amount of data available for cybercrime investigations are growing at an unprecedented rate, creating problems and huge challenges for law enforcement personnel. In order to present digital data as evidence in court, all data sets in the crime scene or laboratory must be examined in detail. Jain et al. [24] proposed the network forensic framework which combines the achievements of the past twenty-five forensic models to create a mechanism for

604

M. Thomas and B. Meshram

creating a new digital forensic model. The planned model offers a consistent investigation technique, which can directly transform model theory into an instrument, an object of historical perspective, reducing the cost and time to investigate any digital crime. Mouhtaropoulos et al. [25] have studied Digital forensics primarily attempts to respond to an information security event.

3 Systematic Networkt Forensic Investigation Model (SNFIM) The planned model will be based on the Network features and the common application vulnerable to the attacks. There are six main stages in this model, the structure specifies in Figs. 3 and 4. Preparation Phase: The initial understanding of the problem, as well as the appropriate tools, are all part of the preparation phase. This step is used to secure authorization and approval, before developing a suitable plan. The planning period is all the work and activity that desires to be done until a real exploration is possible [26]. This includes reviewing forensic rules and guiding principle, obtains search warrants, supporting organization, preparation, and developing suitable strategies and equipment. Acquisition and preservation phase: Prior to the actual forensic examination, the scope of investigating artefacts has to be understood. This phase will clarify which evidences are available and what do we need to acquire in addition. Acquiring and maintaining a clear life cycle and, identifying and collecting evidence of crime Fig. 3 Network forensic investigation phases

A Brief Review of Network Forensics Process Models and a Proposed …

605

Fig. 4 Flow of the network forensic investigation process in the proposed model

includes labelling, packaging, transportation, image capture, and storage of evidence [27]. Therefore, all objects including access permissions, people of the organization, network configurations during the attack, must be lawfully obtained and properly documented in accordance with clear rules. Examination and analysis phase: This is the phase where experts and professionals search for digital evidence by studying and analyzing various digital contents. Legally confiscated and properly protected equipments, its details, technical procedures, acceptable guidelines, and approved forensic tools are used to determine the origin of the crime and also to locate the culprit [28]. The evidence presented depends on the purpose of the appointment or it may contradict the initial theory. Information sharing phase: The easiness with which Internet can be handled, are used effectively on social networks and by the hacker community. The information sharing phase has the ability to obtain a complete criminal profile of an individual, and to develop an effective interrogation strategy. This kind of collaboration and information sharing can effectively contribute to successful events [29].

606

M. Thomas and B. Meshram

Presentation Phase: The results of the examination and analysis are compiled and submitted to the concerned law enforcement authorities of multiple locations. For example. to make sure that the evidence is correctly identified and obtained, the evidence is objectively and adequately protected and the jurisdiction is appropriate, The language used in the presentation be concise and uncomplicated to understand by a judge. As an important step in this process, it is important to remember that the panel of defendants must submit the findings to the court for conciliation and approval before the trial judge or arbitral tribunal [30]. Review Phase: Evaluates the entire investigation and identifies areas for improvement. From the launch of the experiment to the test, the intermediate results are used for future improvements. The experience gained and the lessons learned are used to prepare new process steps and standards. Cases are categorized according to their status and understanding, the case is suspended, on going, and completed. This is done to resolve future complications, such as appealing to the court, appearing or referring to an acquittal [31].

4 Classification of Internet Forensic Investigation Technologies Digital forensic technology on networks is alienated into the following [32, 33]: • • • •

Digital evidence compilation techniques, Digital evidence analysis strategies, Digital evidence retrieval techniques, and Digital evidence documentation technique.

The technology examines and analyzes digital devices, to identify potential sources, transfer electronic data, analyze, search, and protect stored information [34]. All current and accepted network forensic investigations are classified as: 1. 2. 3. 4.

Intrusion detection systems Traceback systems Distributive systems Attack graph systems.

4.1 Intrusion Detection Systems of Internet Forensic Investigation Intrusion detection system (IDS) is software or hardware set-up that monitors the activities of the organization and network for criminal and malicious activities and generates reports for the central administration may be a policy. Figure 5 shows

A Brief Review of Network Forensics Process Models and a Proposed …

607

Fig. 5 Automated digital forensic technique with the intrusion detection

the digital forensic technique with intrusion detection system. The main purpose of infiltration detection systems is to detect events, record information about them, and report attempts [35].

4.2 Traceback in Internet Forensic Investigation It is essential to offer a network expertise IP tracking model based on the assumption and needs listed in the formalities of the process model. The network architecture as shown in Fig. 6, is based on the deterministic packet marking (DPM) using the Autonomous System (AS) approach. In a dual tracking system where the first step includes the AS of each pocket defined by the first column router, and the second step indicates each packet that the AS edge router (ASER) indicates. In both cases, after the pocket is marked, the other router will not be able to mark it. Not every outgoing pocket is marked and incoming pockets are not marked [36].

4.3 Distributive Internet Forensic Investigation The goal of digital forensics is to answer inquisitive and interesting questions: who, how, what, why, when, and where. The entire forensic examination process is implemented using analytical methods such as the Integrated All-in-Encase Selectors Guide, the U.S. Digital Judicial Forensic Analysis Method, and/or the U.S. National

608

M. Thomas and B. Meshram

Fig. 6 Architecture for IP traceback

Justice Analysis Guidelines. The use of an analytics framework in conjunction with a subset of data has the potential to quickly identify information from electronic sources and provide a full backup of the media for analysis [37]. In this study, they suggest a method of rapid forensic analysis of digital forensics, which allows digital forensic data subgroups to periodically review and analyze information contained throughout the forensics.

A Brief Review of Network Forensics Process Models and a Proposed …

609

4.4 Attack Graph of Internet Forensic Investigation To investigate integrated nature and anti-crime attacks, innovative and innovative approach such as bother maps should be used. Harass maps are distinct as a tool for evaluating the performance of bother scenes with the help of acknowledged damages and configurations. Attack graphs are generally used by mainframe administrators and investigators to analyze the type of attack, diverse types of attack methods, detection, and anticipatory channel. Used against these attacks. The different tools in use for generating attack graphs are: • TVA (Topological analysis of network attack vulnerability)—It creates attack maps using a map investigate algorithm. It uses pro maps to produce before and after circumstances. • NETSPA (Network security planning architecture)—Framework uses known vulnerabilities and network security rules to create network models. It serves as a foundation for generational attack mapping to recognize possible attack and root directions. • MULVAL (Multi-host, multistage vulnerability analysis)—Data recording is a structure for integrating damage and system configurations used as a language. It contains a scanner and an analyzer. A logical machine uses data recording rules to record computer behaviour [38].

5 Methodology Digital forensics has straight requirements in the field of Internet and cyber activities, i.e., to prevent further adverse events, successfully identify the events that lead to the crime and identify the culprits. The methodology should improve the existing preventive measures to prevent recurrences as corporate security professionals improve their standards to protect their corporate networks. The rest of the study pattern is summarized in the following sections.

5.1 Research Questions This section describes the background of this research and identifies key research questions and methods for clarifying and defining research focus based on the results of previous attempts and new expectations. • What is the present measurement methods for Netork Forensics tool quality? • What are the changing technique and directions (data related)? • Whether the data remain unaffected (to a level acceptable in a court of law)? The addition and subtraction criteria of the research papers are as follows.

610

M. Thomas and B. Meshram

5.2 Quality Assessment When evaluating the resources, quality assessment (QA) is first used to identify data obtained in meta-analysis, which is very relevant for the resource collection strategy. In addition, quality evaluation results are provided as additional selection criteria.

5.3 Data Extraction and Data Synthesis Data on research questions and activities were collected before the formation of relevant research questions. Once the data collection is complete, the analysis and review process will be conducted and the results will be presented for review The collected data are listed based on the key components of their solution.

5.4 State-of-the-Art Internet Forensic Models 5.4.1

Intrusion Detection System

Achille et al. [39] have studied Abstract layer theory is applied to describe the purpose and objectives of digital forensic equipment. With compression layers, we determine where errors can occur on the equipment and offer tools to protect them. Forensic types are also defined in terms of abstract layers. Abstract layers are not a new concept, but their use in digital forensics has not been properly documented. Barhate et al. [40] studied that the Intrusion Detection System (IDS) is used to determine infiltration. Its main meaning is to attack and react in a judicious approach. In earlier language, IDS activity is for recognition and rejoinder only. IDS were not capable to achieve system status during the infiltration. Therefore, he could not argue the evidence from the attack in its original form. Rani et al. [41] studied that the Cloud computing has become the flagship service model of computer platforms, providing the necessary resources to all types of users. Singhal et al. [42] Data processing and storage technology were explored to better understand the presentation and use of IDS. Current IDS does not support historical data analysis and data summary. This data model was used to analyze network security and detect attacks on services. Analysts can use their data model to manage multiple data sources (e.g. firewall logs, computer calls, network flow data) and to run the required response time for two orders faster than modern technology. Inayat et al. [43] A special feature for accessing cloud resources is the connection link allowing malicious individuals to perform malicious activities at a loss. This protection is a major challenge in the MCC environment [44]. Occupancy processes are studied through the recording and analysis of computer calls. These methods do not provide sufficient detection accuracy because they do not explain internal infiltration events. Sy et al. [45] the aim of this study is

A Brief Review of Network Forensics Process Models and a Proposed …

611

to demonstrate the analytical intrusion detection framework (AIDF) that combines (i) the Probability Model Detection Approach and (ii) the Probability Hypothesis Mechanism that monitors forensic interpretation. Infiltration detection warnings, but unregistered signature rules not expressed in the probability model. AITF compatibility for data integration, (i) understanding of the distribution IDS environment with multiple sensors, and (ii) the mechanism for selecting and integrating multiple model hypotheses. Definition of the most probable forensic examination Zhang et al. [46] studied the Network Intrusion Detection System (NIDS) as an important security tool for detecting malicious activity on modern networks. To speed up the detection of minority classes and ensure performance, proposing a new class of unbalanced processing technology called SGM, a large-scale database that synthetic minority over-sampling technique (SMOTE) and Cassian clustering models. SGM-CNN integrates with a neural network, modifies the unbalanced class process, and investigates the influence of different numbers of nuclei and different learning speeds on model performance. Rahouma et al. [47] discusses the future role of machine learning technology in identifying security threats, the automated response to address security challenges in the Egyptian fiber optic network, and the proposed new threat and response detection model for protecting customer data. Aloqaily et al. [48] ıntroduce an automated secure continuous cloud service access framework for smart connected vehicles that is protected from security attacks and meets service quality (QoE) requirements. Continuous service is available in clusters based on smart car service. Trusted Third Party Companies (TTPs) select cluster leaders who act as intermediaries between service complaints and providers. Customers get better service from selected service providers. Houmansadr et al. [49] reading a mobile device called a smart phone is becoming more and more difficult and powerful to effectively enhance more features, which raises security concerns for smartphone users. Yampolskiy et al. [50] studied human computer communication based on the skills, style, preferences, knowledge or strategy that people use when working with a computer. It analyzes the features described to describe HCI behaviour. Tthe experimental results are presented to determine behavioural interactions in online gaming networks based on the strategies used by players. Zhang et al. [51] proposed lightweight semantic-based knowledge fusion model (SKFM) model to provide natural language processing (NLP) and credit resources using Schools educational documents to generate information. SKFM relies on NLP based on automated components, and initiates fan searches when unsolicited cases occur. Vasudevan et al. [52] network infiltration detection system is an important task to detect the attack. At the time of signature or rule based diagnosis, the previous attack is modelled and signatures/rules are adopted. These rules will be used to detect such attacks in the future, but if there is a malfunction or external detection system, normal network traffic is the standard. Data mining and machine learning techniques are widely used in offline NDS. The characteristics of supervised and supervised training include identification of data templates, identification of publishers, definition of

612

M. Thomas and B. Meshram

learned activity for input characteristics, and generalization of data events. UlltveitMoe et al. [53] provides a mobile intrusion prevention system (mIPS) framework that provides enhanced privacy integrated with the Managed Security Service (MS).

5.4.2

Trace Back System

Armoogum et al. [54] studied Network forensic technology is a new approach with active and real-time response features that can be used to investigate network attack such as denial of service (DoS) attacks. After a compromise, there is always a check and investigation after the attack, so useful immediate evidence is lost. Adjustment is required without the loss of serious traffic and live inquiry. IP tracking schemes can be used to identify the source of the attack. We offer the use of mobile agents for live IP trading. Cheng et al. [55] studied the Cloud services offer great options for the practical use of the IP tracking system. An innovative cloud based tracking framework with many positive features that enable ISPs to use tracking services on their networks. While this will give access to advanced training services, restricting access will become a major issue. Selamat et al. [56] in general, the purpose of all methods is to establish the origin of the incident and to maintain the network of custody so that the course of the legal process can be adopted. The trace-qualification process has become an important or integral part of the digital quest process because it has the ability to generate a map from a variety of sources to obtain event evidence to use event events for other related query characteristics. The purpose of this integration is to provide a link between the evidence, the companies, and the evidence involved in this process, as well as the evidence-based collection and evidence, especially during the digital forensic framework collection phase. Alex et al. [57] due to the popularity and constant availability of the cloud, the issue of security is still a major issue [58]. The user does not know where the data is stored and in which data centre. The distributed nature of the cloud is again the protection point of the cloud, which allows harmful actions to be performed very easily. Although much has been done to ensure the required data protection in Cloud, much work needs to be done to maintain technological advancement, and data production from various computing environments will increase rapidly [59]. Digital Forensics process depends on the continuous TF steps, each step depends on the continuous TF procedures, and each process depends on the assignments and sub-assignments, respectively. There are several TF process models in the literature that define DF stages, but there are no DF models that define step-by-step sequencing procedures for identified crimes [60]. An analytical crime scene procedure model (ACSPM) addresses this shortcoming. In addition to the entire digital investigation process and the completion stages of the trial, crime is the focus of digital case law. Helix, Blackrock Network Security, GNU/Linux A unique solution is to create a unique live CD GNU/Linux distribution after the digital forensic distribution. The resulting software enables effective network monitoring and analysis, further possibilities for future needs, and greater portability compared to standard digital forensic software [61].

A Brief Review of Network Forensics Process Models and a Proposed …

613

Kwon et al. [62] the leader of the global joint investigation was arrested in March 2018, but in the event of a Bank of Bangladesh robbery, the damages have not yet been resolved due to a lack of digital forensic evidence. This study offers the best IP forensic tracking and visualization techniques for digital forensics. Various process models describing digital court procedures and procedures have been developed with issues associated with these models [63]. Security incidents targeting information systems as complex and intricate items may be investigated and incoming individuals may be responsible for the lack of evidence of guilt. They have set up a digital forensic system (DigFornet) [64] on the net that can be used to analyze security incidents and explain the activities of attackers. The Digfornet Infiltration Response Team uses knowledge and system tools to reconstruct attack potential and show how a computer works at each stage of the scene and introduced a hypothetical concept to address the lack of data related to attack scenes or automated and lost evidence of investigative knowledge. Industrial Revolution 4.0 [65] explores the automatic communication of digital devices around the world, including cyber physics devices, IoT devices, mobile devices, storage devices, and many other digital devices, including PCs and PCs; The rate of cyber crime will increase. It questions the need for an additional digital forensics investigation framework (DFIF) to effectively investigate digital crimes in court; the integrity of the resources must be maintained when working on the structure. Mapping leads to a better investigation process. Jain et al. [66] the use of technology is important. Digital forensics is a specialized field of computer forensics in which scientific procedures and tools are used to obtain digital evidence in court. The main goal is to introduce a digital forensic tool based on the digital forensic framework. Wazid et al. [67] clarified the biggest challenge in the cyber world has emerged as realism. Many digital forensic devices are being developed to meet this challenge, but hackers are developing counter-technologies at the same pace. It covers the basics of digital expertise along with the latest trends in social networking, cloud computing, websites and phishing.

5.4.3

Distributed System

Trcek et al. [68] it turned out those global networks were moving towards serviceoriented structures and touch networks. On the one hand, Spectrum creates a mix of new services with multiple services, and on the other hand, it represents the influence of the Internet. These two types are different and can be used in many different ways, which can lead to serious complications in case of abuse. As a result, both trends increase the need for new approaches to digital skills, as evidenced. Since technology alone is not enough, it should provide adequate support for relevant investigative procedures that have become the subject of international consensus. Singh et al. [69] the computer system reads forensic tracking and these files contain unstructured text, so their analysis is difficult, leading to the introduction of automated technology that provides new and important information on how to display the clustering of displayed documents. This can be achieved by changing the algorithm differently from the content used to create the cluster. They work here in the expert process

614

M. Thomas and B. Meshram

of document summary in expert investigation. Here we associate it with the Kalgorithm, which improves the performance of computers and actively converts this work using MATLAB. Roussev et al. [70] distributed digital forensic (DDF) devices and traditional intelligence tools operating in a single station offer numerous real-world examples of their range, which is a serious barrier to the timely processing of digital evidence. Based on the observations of routine intelligence work, we will highlight the DDF software about computer requirements. Describe another modern forensic technology implemented by switching to DDF equipment. Ellison et al. [71] it has been explored that there are different models in the areas of digital expertise and security due to the unique function of each knowledge. A knowledge base that operates on a set of assumptions, facts, and laws is called a knowledge base, and is built on the structure of dynamics. Ontology has its philosophical origins; it deals with existence and what is true. In this case, gynaecology, which is presented as the basis of knowledge, requires the suggestions (and universes) of truth based on knowledge and logic. Such clear ideas are expressed as explanatory logic, so they are logically useful. However, existing digital forensic logic and knowledge management models do not contain clear principles and concepts that can be used to build a general knowledge base for digital forensics and security. To solve this study, important theories and hypotheses have been developed that include explanatory logic that can be used reliably. Satpathy et al. [72] investigate whether security has become a major issue in cyberspace. Digital analysis of forensics at a relatively young age has become an unstoppable victim of the rapid advancement of computer technology, so this is possible thanks to new and innovative computer approaches. Digital forensic analysis is unique to forensics because it contains more data from the investigation than mathematical and other types of forensic problems. The digital investigation process can be performed using several forensic models. The Domain Operations Panel of Data Integration reflects on balancing the investigative process, creating high quality data for analysis, and generating legal digital evidence as a court certification expert. Beebe et al. [73] digital text string search tools use appropriate and/or indexing methods to identify physical sources at the physical level to search for the given text strings. They are designed to provide 100% on-demand memory (e.g. find all examples of text strings). Depending on the nature of the data set, this leads to very high success rates and may not meet demand targets. When websites are similar to search engines, they use a ranking algorithm to get the most efficient and effective search results. This is the reason why digital forensic text search tools fail to determine team and/or search success, which significantly improves (or accelerates) the ability of searchers to succeed. Beebe et al. [74] investigators and researchers have found that large and terabitsized data sets are increasingly being used to conduct digital surveys. Sophisticated digital search tools and processes control work from a computer and human perspective because they constantly rely on very simple algorithms to reduce and recover data. Expanding data mining research in the field of digital forensics would have the following or several advantages: (i) reduce data processing and human processing

A Brief Review of Network Forensics Process Models and a Proposed …

615

time; (ii) improved quality of data related to data analysis; (iii) Reduce the cost associated with digital claims. It offers recommendations for using digital forensics for data forensics. Bermad et al. [75] the effectiveness of the mobile investigation process (smartphone forensics) is related to its evidence analysis phase. Depending on this collection and location of all resources, their temporary, functional and associated combinations. The abundance of evidence, their complexity, and the number of correlations between different types of data complicate the analysis phase of the evidence and the reconstruction of the crime. They provide a method of temporal and functional analysis based on data processing (supervisory classification). Introduce new clustering techniques based on dynamic causes and event reconstruction (SMS and calls). In this case, we can help the investigator to identify inconsistencies and information about the crime and to inform everyone of the global outlook on all events. Evidence was collected. Verma et al. [76] rapid technological changes, the use of digital tools (diversity and numbers), and the large amount of data available on these devices have been found to continue to challenge the current state of digital expertise. Because the protection of confidentiality and the completeness of the investigation are mutually exclusive, digital forensic testing focuses on the effectiveness of the investigative process or the protection of data confidentiality to address these challenges. However, the general approach to data confidentiality remains an open question that does not affect the investigator’s capabilities or the overall performance of the intelligence process. Duce et al. [77] examined the rapid emergence of technological development and the challenges of digital forensics in the face of attacks on criminal thinkers that are naturally distributed in natural development. Collect evidence and make assumptions at the Digital Forensic Centre. Marangos et al. [78] CC studied the impact of the model on digital forensic systems. Digital forensics (DF) is a field of computer science that generates relevant evidence in the digital form acceptable and scientifically proven for use in civil or criminal proceedings. Their analysis significantly affects the right to share digital resources inherited from the CC model. Homem et al. [79] the mobile devices and virtual devices in the Internet of Things will gradually become the endpoints of different networks owned by different parties, while engaging in unlicensed licensing or illegal activities. Digital Forensics and Incident Response (DFIR) tools today are struggling to conduct digital research in a freely controlled online environment because they have a number of issues: lack of resources, availability, reliability, confidentiality, data size, speed, and diversity. They integrate the structural and structural P2B overlays of our architecture to enable efficient capture, analysis. Singh et al. [80] security vulnerabilities in Windows operating systems that have been studied for years have increased significantly, and the most serious threat is malware (malware). Theoretically, the recovery cache stores the unstable memory of the attack process and the affected machine in the Windows registry. Digital Forensic Product Process (DFR) offers suggestions for collecting active digital tracks. Tewelde et al. [81] the concept of “scientific theory” is even more important in the new field of digital forensics, because not all empiricism is hypothetical. Although

616

M. Thomas and B. Meshram

the term “concept” is widely used in digital expert narrative, its use is not adequately reflected from a scientific point of analysis. This issue is discussed with special reference to the proper functioning of the carrier where concept plays an important role. Bashir et al. [82] with the regular use of the Internet and the advancement of technology, cyber and malware applications are growing on digital devices. Activities performed electronically can be analyzed using digital forensic methods.

5.4.4

Attack Graph System

Wang et al. [83] proposed a network prototype a skill that integrates the automatic logic of presentation, fabrication, and infiltration evidence. We offer a new map of Evidence Map to help you navigate and process navigation resources. For automated analysis of evidence, we develop a hierarchical basic framework that integrates local logic with global logic. In local logic, we use Rule-based Fuzzy Cognitive Maps (RBFCM) to model the evolution of the status of suspicious hosts. In order to achieve a global goal, we aim to identify the host group that is strongly associated with the attack and gain their connection in the context of the attack. Zhu et al. [84] it has been learned that remote hackers rarely seize network attacks on computers; When an attack is detected successfully or simply, it is often advisable to reconstruct the background of this security breach: all violations and related incidents. They record the latest network traffic data to identify these instances of attack. Their simulations show that the algorithm achieves accuracy in detecting attack patterns. The added benefit of not tracking their attack mode detection, for example, is that it does not require a user-defined gateway. Barik et al. [85] As computer networks evolve day by day, network security has become a major issue. At the same time, attacks become more complex and computer networks become more secure. Attack Map is a modelling tool used to evaluate the security of organizational charts. Since its introduction, significant research efforts have been made to develop a theory and application around the concept of an attack map. Liu et al. [86] it explores how difficult it is to model an effective and accurate security event to determine the attack situation for an enterprise network. In this study we discuss how to create an attack situation and source map using the evidence of security studies. To achieve the accuracy and completeness of the source map, we use introductory induction and hijacking logic to verify the evidence. In addition, as the structural circumstances and the evidence presented may stand in court, federal evidence laws will be taken into account to determine the evidence in advance. Barrère et al. [87] the network has been explored as an effective mechanism for guiding investigators during forensics. To this end, they present the concept of a basic offense map, a brief overview of key pathways that can attack targets on a particular network. Such a summary allow forensic investigators to spotlight on important nodes that may be part of an attack path, thereby minimizing common nodes (devices, network privileges). Conlan et al. [88] Expert anti-expert tools, techniques, and methods were studied, which became a strong barrier to the digital forensic community. Strategies need to be developed to address this growing problem. Al-Mahrouqi et al. [89] achieving high security on

A Brief Review of Network Forensics Process Models and a Proposed …

617

low-end and wired communication networks is consistently difficult to achieve. The growing rate of secure network owners and attackers is relatively close. Ha et al. [90] studied that the capability acquisition graphs (CAGs) that provide a robust framework for modelling internal threats and attacks, and the system vulnerabilities. The CAG-based security modelling systems have not yet been used in practice. In particular, it describes an information-based graphical database tool known as ICMP. Imran et al. [91] the growing nature and increasing demand for the open source cloud is making the infrastructure an ideal target for malicious attacks, unauthorized access to data storage, and a serious threat to cloud software security. In the event of any malicious action, the cloud resource information used by digital forensics to diagnose the problem may fall prey to malicious companies and cloud security software. Liu et al. [92]. Problems with reconstruction of attack scenes include large amounts of data is damaged or destroyed. This model solves these problems using a variety of methods, including evidence-based displays, including computer damage, inductive logic, and reconstruction of attack scenes. In the introduction the system uses known vulnerability databases and expert anti-databases, which extend to standard databases such as the NSD National Vulnerability Database. Created for network criminal analysis, this system reduces the time and effort required to reach final conclusions about network attacks. Peisert et al. [93] Judgments in the field of computer forensics were often considered conditional. It discusses the need for rigorous expertise and outlines the features that should exist in this model. This requires far less carefully selected and effective data than the forensic model. Poolsapassit et al. [94] the computer log files contained vital evidence related to the computer attack. However, journal files are often so large that most of the information in them is irrelevant to the query. Rekhis et al. [95] security measures and the logic of digital inquiry were tested in its high-level specification language. Rekhis et al. [96] studied the analysis and reconstruction of attack processes, and motivates attackers to resist counter-attacks, whether it is difficult, difficult or impossible to defeat the investigative process. The literature has suggested several methods for systematically reconstructing the development of actions that occur during an event use theoretically and scientifically proven methods. However, these methods are not intended to address forensic attacks because they believe that the evidence collected is reliable and that experts have not modelled the operations and have developed samples of attacks, security solutions and resources. Liu et al. [96] learned that attack maps are used to estimate attack paths and known computer damage from a computer configuration. Attack maps can be used to determine the order of damage used to carry out attacks, and forensics can help identify potential attacks.

618

M. Thomas and B. Meshram

6 Comparative Analysis of State-of-Art Models 6.1 Intrusion Detection System in Network Forensic Analysis In this subsection, we analyze the different IDS for digital forensic analysis. It is very irrational to find infiltration as a discipline. Table 1 shows the brief review on IDS for digital forensic analysis. Form state-of-art papers, most of the studies discussed the intrusion detection through attack types which are DoS, DDoS, botnet, portnet and web attack. Table 1 IDS in forensic analysis IDS

IP spoofing attacks

SNORT

Accurate authentic

[39]

IDS

DOS: port scan: DDoS

SNORT

Nfinimize false alarm rate

[40]

NIDS

Cloud anti forensic

SNORT

RBNN+, Kmeans+, Correlation

IDS

DOS Attack

SNORT

NetSTAT- real time network: smurfand SYN flood

[42]

DRS

Dos and DDoS

SNORT

Multi-tenancy and distributed neÄvork feature

[43]

PIDAS

Botnet attack

SMOTE

Potential system vulnerabilities

[44]

AIDE and IDS

DOS attack

SNORT

Augnlented feature and speculative fast

[45]

NIDS

Dos, DDoS, Bomet: Port Scan, Web attack

SMOTE

K-means homogeneity metri feature GBT, DNN, binary and multiclassification

[46]

IDS

Dos, DDoS

SNORT

Artificial intelligence (Al) prediction feature

[47]

IDS

Port Scan

SNORT

NS-3 collected

[48]

IDS

Dos, DDoS

SNORT

Kmeans, ANN

[49]

GUI

Web attack

SMOTE

Time between windows and new windows

[50]

Dos, DDoS

SMOTE

Multi-tenancy

[51]

NIDS

DOS attack

SNORT

SSENet

[52]

mlRS, IDS and IPS

DOS and DDoS attack

SNORT

No. of TCP/IP connective [53] CPU usage, battery cllarge

A Brief Review of Network Forensics Process Models and a Proposed …

619

Table 2 Trace back techniques in forensic analysis Trace-back scheme

Advantages

Disadvantages

Sources

Input debugging

Support Incremental Implementation Can be used against both DOS or DDoS Functions are not necessary to perform

Utilization of time is high DOS or DDoS attack at the same time For a successful trace: the attack must last a long time

[54, 55]

Controlled flooding

Easy to implement Can be used against both DOS or DDoS Suitable for network infrastructure

Utilization of time is high [56–58] Cannot distinguish DDoS and genuine flash crowed It can only be used in attacks

Logging

Easy to implement Using single packet to reconstruct attack path Allowing post packet analysis

Have potential hash collision [59–61] Substantial storage required Vulnerable to compromised router

Hop count filtering

Easy to implementation Bandwidth above is very low Less infrastructure support

Long detection cycle Lack of economic incentive Vulnerable to compromised router

[62–64]

Packet marking

It is scalable Applying the program is Low processing very expensive Suitable for a variety of attacks Costing data fragmentation High bandwidth overhead

[65–67]

6.2 Tracebacks in Network Forensic Analysis In this section, we discuss the development of effective forensic surveillance technologies for dealing with cybercrime through anonymous communication networks. Associations around the world today rely heavily on the Internet for business, military communications, and information dissemination. However, this has led to cyber security issues. Network expertise plays an important role in supporting legal oversight as the number of cybercrimes increases in the integrated and fast-growing Internet world. Table 2 describes the trackbacks usage in Network Forensic analysis.

6.3 Distributive Systems in Forensic Analysis Digital forensics is a bough of forensics that deals with the examination of digital crime, which is primarily an attack on digital plans. There are several subdivisions of digital forensics based on objectives, counting computer forensics, network forensics, and mobile forensics. The aptitude to conduct digital research is straight connected to the accessibility of confirmation, chiefly in journal records. It includes databases where researchers can find information about the causes of crime and their source

620

M. Thomas and B. Meshram

(computer, laptop, and tablet). As a result, it causes the attacker to be found and/or confiscated. In the age of the Internet of Things (IoT), digital forensics will have fun an important role in investigate crime alongside IoT devices (Table 3). Table 3 Distributive techniques in forensic analysis Source

Category

Approach

Limitations

[68]

Cognitive mapping

Top down

MACs and digital signature to provide more reliability

[69]

Fuzzy c-means

Hierarchical

Clustering generates subset of clusters or partition

[70]

Gigabit Ethernet

DDF toolkit

DDF toolkit is similar to commercial forensic tools

[71]

Description logics

Ontology based

Language defined for computer reasoning must have rules

[72]

DFRWS framework

Cyberspace based

Comprehension or the need to reduce information overload

[73]

EF database

CPC database

Rules for conducting forensic analysis can be stored in CPC

[74]

Mining algorithm

Social network analysis

Uses of data discrimination like data characterization

[75]

WWW Consortium

Combichem approach

W3C RDF vocabulary description language used

[76]

Cloud computing

RCG based

CSP has no privilege on the upper layer

[77]

Hashing algorithm

Defense in depth

Unstructured P2P overlays in various capacities

[78]

RSA encryption

Trigger-based

RaaS provides a ransomware service

[79]

Notorious Duhem-Quine dilemma

Hypothesis based approach

Carrier’s framework is normative rather than descriptive

[80]

DES algorithm

Live Forensic Analysis

FAT do not support bitlock drive encryption drive

[81]

Clustering algorithm

Kohonen’s self organising map (SOP)

Feasibility at low risk

[82]

CATFA

CATFA approacgh

Scalable with many events because it is parallel

A Brief Review of Network Forensics Process Models and a Proposed …

621

6.4 Attack Graphs in Forensic Analysis Regardless of momentous efforts to look after network from cyber-attacks (Gartner, Inc. 2014), computer administrators have not been able to cope with modern threat technology, as evidenced by the recent history of corporate infringement (Lord 2015). One of the largely ordinary network protection strategies is to detect and insert risks. Such an approach requires network risk assessment, prioritization of the most important threats, and subsequent evaluation of risk performance (Wheeler 2011). Finally, administrators use these values to select the appropriate commands. But most of the time this analysis is implemented individually for each network component, i.e., ignoring the relationship between the vulnerabilities, i.e. how one vulnerability can be successfully exploited so that the attacker can exploit vulnerability and thus gain exclusive rights across the networkas given in Table 4.

7 Conclusion In this study, we presented an undemanding framework for digital inquiry based on the causes and consequences of events. The benefits of this work contribute to the development of a common solution that meets the needs of a swiftly altering and greatly impulsive digital technology environment. Gain the honesty and acceptance of digital resources and internetworking devices. The solution to the research questions is as follows. This paper describes in detail the various techniques used in basic Forensic Methodologies with comparisons on all relevant strategies. Network forensics can be classified is based on four technologies: infiltration detection, tracking, distribution, and attack maps.

622

M. Thomas and B. Meshram

Table 4 Attack graphs in forensic analysis Source

Category

Technique

Vulnerability

Effect

[83]

Attack tool

Tool

Clustering technique

Unix

Attack instances are propagated

[84]

Attack tool

RBFCM

Buffer overflow

Analytics system effectively integrates feedback into an automated reasoning process

[85]

Attack tool

Obfuscate signature

SNORT rule

Mislead forensic analysis by crafting image

[86]

Destroy data

BCWipe

Remove log file

CVE-2009-2446

Delete data permanently

[87]

Attack tool

Wireshark

SQL injection attack

GNS3

Was designed to increase the filter-ing process of the output event logs

[88]

Attack tool

Nu SMV tool

ICMAP

CERT

Visualization used as an effective method for integrating and analysing diverse information

[89]

Destroy data

BCWipe

Delete content

CVE-2009-1918

Access control modules

[90]

Attack tool

SQL injection

Buffer overflow

Eliminates info from log file

[91]

Attack tool

Steganography

SNORT rule

Provenance detection

[92]

Attack tool

Steganography

Buffer overflow

Bypass being detected by rules

[93]

Attack tool

Postgres SQL

Linux

[94]

Attack tool

AI techniques

Unix

System dependent

[95]

Attack tool

I-TLA

SNORT rule

Bypass being detected by rules

[96]

Attack tool

SQL injection

SNORT rule

Attacker could alter the evidence

[96]

Attack tool

Obfuscate signature

SNORT rule

Bypass being detected by rules

CVSS

Appraiser tool

Model checker

A Brief Review of Network Forensics Process Models and a Proposed …

623

References 1. Pilli ES, Joshi RC, Niyogi R (2010) Network forensic frameworks: survey and research challenges. Digit Investig 7(1–2):14–27 2. Mohamed IA, Manaf ABA (2014) An enhancement of traceability model based-on scenario for digital forensic investigation process. In: 2014 third international conference on cyber security, cyber warfare and digital forensic (CyberSec). IEEE, pp 12–15 3. Palmer GL (2001) A road map for digital forensic research. Technical Report DTR-T0010-01, DFRWS. Report for the First Digital Forensic Research Workshop (DFRWS) 4. Sivaganesan DD (2021) A data driven trust mechanism based on blockchain in IoT sensor networks for detection and mitigation of attacks. J Trends Comput Sci Smart Technol 3(1):59– 69. https://doi.org/10.36548/jtcsst.2021.1.006 5. Cleland-Huang J, Hayes JH, Domel JM (2009) Model-based traceability. In: Proceedings of the 2009 ICSE workshop on traceability in emerging forms of software engineering. IEEE Computer Society, pp 6–10 6. Carrier B, Spafford E (2004) An event-based digital forensic investigation framework. Digit Investig 7. Agarwal A, Gupta M, Gupta S, Gupta SC (2011) Systematic digital forensic investigation model. Int J Comput Sci Secur (IJCSS) 5(1):118–131 8. Chung H, Park J, Lee S, Kang C (2012) Digital forensic investigation of cloud storage services. Digit Investig 9(2):81–95 9. Köhn M, Olivier MS, Eloff JH (2006) Framework for a digital forensic ınvestigation. In: ISSA, pp 1–7 10. Perumal S (2009) Digital forensic model based on Malaysian investigation process. Int J Comput Sci Netw Secur 9(8):38–44 11. Ademu IO, Imafidon CO, Preston DS (2011) A new approach of digital forensic model for digital forensic investigation. Int J Adv Comput Sci Appl 2(12):175–178 12. Delport W, Köhn M, Olivier MS (2011) Isolating a cloud instance for a digital forensic investigation. In: ISSA 13. Perumal S, Norwawi NM, Raman V (2015) Internet of Things (IoT) digital forensic investigation model: top-down forensic approach methodology. In: 2015 fifth ınternational conference on digital ınformation processing and communications (ICDIPC). IEEE, pp 19–23 14. Valjarevic A, Venter HS (2012) Harmonised digital forensic investigation process model. In: 2012 ınformation security for South Africa. IEEE, pp. 1–10 15. Kanellis P, Kiountouzis E, Kolokotronis N, Martakos D (2006) Digital crime and forensic science in cyberspace. Idea Group Publishing, London, UK 16. Kebande VR, Ray I (2016) A generic digital forensic investigation framework for internet of things (IoT). In: 2016 IEEE 4th ınternational conference on future ınternet of things and cloud (FiCloud). IEEE, pp 356–362 17. Sathwara S, Dutta N, Pricop E (2018) IoT forensic a digital investigation framework for IoT systems. In: 2018 10th ınternational conference on electronics, computers and artificial ıntelligence (ECAI). IEEE, pp 1–4 18. Shrivastava G, Gupta BB (2014) An encapsulated approach of forensic model for digital investigation. In: 2014 IEEE 3rd global conference on consumer electronics (GCCE). IEEE, pp 280–284 19. Valjarevic A, Venter HS, Ingles M (2014) Towards a prototype for guidance and implementation of a standardized digital forensic investigation process. In: 2014 ınformation security for South Africa. IEEE, pp 1–8 20. Omeleze S, Venter HS (2013) Testing the harmonised digital forensic investigation process model-using an Android mobile phone. In: 2013 ınformation security for South Africa. IEEE, pp 1–8 21. Kebande VR, Karie NM, Venter HS (2016) A generic digital forensic readiness model for BYOD using honeypot technology. In: 2016 IST-Africa week conference. IEEE, pp 1–12

624

M. Thomas and B. Meshram

22. Burrows C, Zadeh PB (2016) A mobile forensic investigation into steganography. In: 2016 international conference on cyber security and protection of digital services (Cyber Security). IEEE, pp 1–2 23. Charles T, Pollock M (2015) Digital forensic investigations at universities in South Africa. In: 2015 second ınternational conference on ınformation security and cyber forensics (InfoSec). IEEE, pp 53–58 24. Kao DY, Wu NC, Tsai F (2019) The governance of digital forensic investigation in law enforcement agencies. In: 2019 21st ınternational conference on advanced communication technology (ICACT). IEEE, pp 61–65 25. Jain N, Kalbande DR (2015) Digital forensic framework using feedback and case history keeper. In: 2015 ınternational conference on communication, ınformation & computing technology (ICCICT). IEEE, pp 1–6 26. Mouhtaropoulos A, Grobler M, Li CT (2011) Digital forensic readiness: an insight into governmental and academic initiatives. In: 2011 European ıntelligence and security ınformatics conference. IEEE, pp 191–196 27. Kyei K, Zavarsky P, Lindskog D, Ruhl R (2012) A review and comparative study of digital forensic investigation models. In: International conference on digital forensics and cyber crime. Springer, Berlin, Heidelberg, pp 314–327 28. Boateng R et al (2010) Cyber crime and criminality in Ghana: its forms and implications. In: Proceedings of the 16th Americas conference on ınformation systems 29. Smith RG, Grabosky PN, Urbas G (2004) Cybercriminals on trial. Cambridge University Press. ISBN: 9780521840477 30. Kent K, Chevalier S, Grance T, Dang H (2006) NIST SP 800-86 guide to ıntegrating forensic techniques into ıncident response 31. Bunting S (2007) Mastering Windows network forensic and investigation, 1st edn. Sybex. ISBN-13: 978-0470097625 32. Volonino L, Anzaldua R, Godwin J (2007) Computer forensics: principles and practices. Pearson Prentice Hall, NJ, USA 33. Jiao F, Gao W, Duan L, Cui G (2001) Detecting adult ımage using multiple features. In: Proceedings of the ICII 2001, vol 3, pp 378–383 34. Phung S, Chai D, Bouzerdoum A (2003) Adaptive skin segmentation in color ımages. In: IEEE international conference on acoustics, speech, and signal processing (ICASSP 2003) 35. Cosic J, Baca M (2010) A framework to (im)prove “chain of custody” in digital ınvestigation process. In: Proceedings of the CECIIS, Varazdin, Croatia 36. Hemdan EED, Manjaiah DH (2018) Cybercrimes investigation and intrusion detection in internet of things based on data science methods. In: Cognitive computing for big data systems over IoT. Springer, Cham, pp 39–62 37. Pilli ES, Joshi RC, Niyogi R (2010) An IP traceback model for network forensics. In: International conference on digital forensics and cyber crime. Springer, Berlin, Heidelberg, pp 129–136 38. Quick D, Choo KKR (2017) Big forensic data management in heterogeneous distributed systems: quick analysis of multimedia forensic data. Softw: Pract Exp 47(8):1095–1109 39. Chhabra GS, Singh P (2015) Distributed network forensics framework: a systematic review. Int J Comput Appl 119(19) 40. Achille MM, Roger AE (2014) Obtaining digital evidence from intrusion detection systems. Int J Comput Appl 95(12) 41. Barhate K, Jaidhar CD (2013) Automated digital forensic technique with intrusion detection systems. In: 2013 3rd IEEE ınternational advance computing conference (IACC). IEEE, pp 185–189 42. Rani DR, Geethakumari G (2020) A framework for the identification of suspicious packets to detect anti-forensic attacks in the cloud environment. Peer-to-Peer Netw Appl 1–14 43. Singhal A, Jajodia S (2006) Data warehousing and data mining techniques for intrusion detection systems. Distrib Parallel Databases 20(2):149–166

A Brief Review of Network Forensics Process Models and a Proposed …

625

44. Inayat Z, Gani A, Anuar NB, Anwar S, Khan MK (2017) Cloud-based intrusion detection and response system: open research issues, and solutions. Arab J Sci Eng 42(2):399–423 45. Xie Y, Feng D, Tan Z, Zhou J (2016) Unifying intrusion detection and forensic analysis via provenance awareness. Future Gener Comput Syst 61:26–36 46. Sy BK (2009) Integrating intrusion alert information to aid forensic explanation: an analytical intrusion detection framework for distributive IDS. Inf Fusion 10(4):325–341 47. Zhang H, Huang L, Wu CQ, Li Z (2020) An effective convolutional neural network based on SMOTE and Gaussian mixture model for ıntrusion detection in ımbalanced dataset. Comput Netw 107315 48. Rahouma K, Ali A (2019) Applying intrusion detection and response systems for securing the client data signals in the Egyptian optical network. Procedia Comput Sci 163:538–549 49. Aloqaily M, Otoum S, Al Ridhawi I, Jararweh Y (2019) An intrusion detection system for connected vehicles in smart cities. Ad Hoc Netw 90:101842 50. Houmansadr A, Zonouz SA, Berthier R (2011) A cloud-based intrusion detection and response system for mobile phones. In: 2011 IEEE/IFIP 41st international conference on dependable systems and networks workshops (DSN-W). IEEE, pp 31–32 51. Yampolskiy RV (2007) Human computer interaction based intrusion detection. In: Fourth ınternational conference on ınformation technology (ITNG’07). IEEE, pp 837–842 52. Zhang Y, Saberi M, Chang E (2018) A semantic-based knowledge fusion model for solution-oriented information network development: a case study in intrusion detection field. Scientometrics 117(2):857–886 53. Vasudevan AR, Selvakumar S (2016) Local outlier factor and stronger one class classifier based hierarchical model for detection of attacks in network intrusion detection dataset. Front Comp Sci 10(4):755–766 54. Ulltveit-Moe N, Oleshchuk VA, Køien GM (2011) Location-aware mobile intrusion detection with enhanced privacy in a 5G context. Wirel Pers Commun 57(3):317–338 55. Armoogum S, Mohamudally N (2008) Mobile agent based real time IP traceback. In 2008 third ınternational conference on digital ınformation management. IEEE, pp 69–74 56. Cusack B, Tian Z, Kyaw AK (2016) Identifying DOS and DDOS attack origin: IP traceback methods comparison and evaluation for IoT. In: Interoperability, safety and security in IoT. Springer, Cham, pp 127–138 57. Selamat SR, Yusof R, Sahib S, Hassan NH, Abdollah MF, Abidin ZZ (2011) Traceability in digital forensic investigation process. In: 2011 IEEE conference on open systems. IEEE, pp 101–106 58. Alex ME, Kishore R (2017) Forensics framework for cloud computing. Comput Electr Eng 60:193–205 59. Shah JJ, Malik LG (2014) An approach towards digital forensic framework for cloud. In: 2014 IEEE ınternational advance computing conference (IACC). IEEE, pp 798–801 60. Koleoso RA, Ajayi OO, Oladeji FA, Uwadia CO (2018) A digital forensics investigatıon model that incorporates digital chain of custody for its integrıty, and authenticity. Computer Networks, Infrastructure Management and Securıty (CoNIMS), p 55 61. Bulbul HI, Yavuzcan HG, Ozel M (2013) Digital forensics: an analytical crime scene procedure model (ACSPM). Forens Sci Int 233(1–3):244–256 62. Chlapoutakis G (2022) Use of network monitoring and analysis tools and methodologies in digital forensic ınvestigations 63. Kwon S, Jeong J, Shon T (2019) Digital forensic readiness for financial network. In: 2019 ınternational conference on platform technology and service (PlatCon). IEEE, pp 1–4 64. Montasari R (2016) An ad hoc detailed review of digital forensic investigation process models. Int J Electron Secur Digit Forens 8(3):205–223 65. Rekhis S, Krichene J, Boudriga N (2008) DigForNet: digital forensic in networking. In: IFIP ınternational ınformation security conference. Springer, Boston, MA, pp 637–651 66. Singh KS, Irfan A, Dayal N (2019) Cyber forensics and comparative analysis of digital forensic ınvestigation frameworks. In: 2019 4th ınternational conference on ınformation systems and computer networks (ISCON). IEEE, pp 584–590

626

M. Thomas and B. Meshram

67. Jain N, Kalbande DR (2014) A comparative study based digital forensic tool: complete automated tool. Int J Forens Comput Sci 68. Wazid M, Katal A, Goudar RH, Rao S (2013) Hacktivism trends, digital forensic tools and challenges: a survey. In: 2013 IEEE conference on ınformation & communication technologies. IEEE, pp 138–144 69. Trˇcek D, Abie H, Skomedal Å, Starc I (2010) Advanced framework for digital forensic technologies and procedures. J Forens Sci 55(6):1471–1480 70. Singh AP, Jain RC (2015) Group identification based classification technique for aggregated common data used in digital forensic ınvestigation (DFI) 71. Roussev V, Richard III G (2004) Breaking the performance wall—the case for distributed digital forensics. Digit Investig 72. Ellison D, Ikuesan AR, Venter H (2019) Description logics and axiom formation for a digital forensics ontology. In: European conference on cyber warfare and security. Academic Conferences International Limited, pp 742-XIII 73. Satpathy S, Pradhan SK, Ray BB (2010) A digital investigation tool based on data fusion in management of cyber security systems. Int J Inf Technol Knowl Manag 2(2):561–565 74. Beebe NL, Clark JG (2007) Digital forensic text string searching: improving information retrieval effectiveness by thematically clustering search results. Digit Investig 4:49–54 75. Beebe N, Clark J (2005) Dealing with terabyte data sets in digital investigations. In: IFIP ınternational conference on digital forensics. Springer, Boston, MA, pp 3–16 76. Bermad N, Kechadi MT (2016) Evidence analysis to basis of clustering: approach based on mobile forensic investigation. In: 2016 7th ınternational conference on sciences of electronics, technologies of ınformation and telecommunications (SETIT). IEEE, pp 300–307 77. Verma R, Govindaraj J, Gupta G (2018) DF 2.0: designing an automated, privacy preserving, and efficient digital forensic framework 78. Duce D, Mitchell F, Turner P (2007) Digital forensics: challenges and opportunities. In: 2nd conference on advances in computer security and forensics (ACSF), LJMU, Liverpool, UK 79. Marangos N, Rizomiliotis P, Mitrou L (2012) Digital forensics in the cloud computing era. In: 2012 IEEE Globecom workshops. IEEE, pp 775–780 80. Homem I, Kanter T, Rahmani R (2016) Improving distributed forensics and incident response in loosely controlled networked environments. Int J Secur Appl 10(1):385–414 81. Singh A, Ikuesan AR, Venter HS (2018) Digital forensic readiness framework for ransomware investigation. In: International conference on digital forensics and cyber crime. Springer, Cham, pp 91–105 82. Tewelde S, Gruner S, Olivier M (2015) Notions of hypothesis in digital forensics. IFIP international conference on digital forensics. Springer, Cham, pp 29–43 83. Bashir MS, Khan MNA (2013) Triage in live digital forensic analysis. Int J Forens Comput Sci 1:35–44 84. Wang W, Daniels TE (2005) Network forensics analysis with evidence graphs (demo proposal). In: Proceedings of the digital forensic research workshop 85. Zhu Y (2011) Attack pattern discovery in forensic investigation of network attacks. IEEE J Sel Areas Commun 29(7):1349–1357 86. Barik MS, Sengupta A, Mazumdar C (2016) Attack graph generation and analysis techniques. Def Sci J 66(6):559 87. Liu C, Singhal A, Wijesekera D (2014) A model towards using evidence from security events for network attack analysis. In: WOSIS, pp 83–95 88. Barrère M, Steiner RV, Mohsen R, Lupu EC (2017) Tracking the bad guys: An efficient forensic methodology to trace multi-step attacks using core attack graphs. In: 2017 13th international conference on network and service management (CNSM). IEEE, pp 1–7 89. Conlan K, Baggili I, Breitinger F (2016) Anti-forensics: furthering digital forensic science through a new extended, granular taxonomy. Digit Investig 18:S66–S75 90. Al-Mahrouqi A, Abdalla S, Kechadi T (2015) Cyberspace forensics readiness and security awareness model. Int J Adv Comput Sci Appl 6:123–127

A Brief Review of Network Forensics Process Models and a Proposed …

627

91. Ha D, Upadhyaya S, Ngo H, Pramanik S, Chinchani R, Mathew S (2007) Insider threat analysis using information-centric modeling. IFIP international conference on digital forensics. Springer, New York, NY, pp 55–73 92. Imran A, Aljawarneh S, Sakib K (2016) Web data amalgamation for security engineering: digital forensic investigation of open source cloud. J UCS 22(4):494–520 93. Liu C, Singhal A, Wijesekera D (2015) A logic-based network forensic model for evidence analysis. IFIP international conference on digital forensics. Springer, Cham, pp 129–145 94. Peisert S, Bishop M, Karin S, Marzullo K (2007) Toward models for forensic analysis. In: Second international workshop on systematic approaches to digital forensic engineering (SADFE’07). IEEE, pp 3–15 95. Poolsapassit N, Ray I (2007) Investigating computer attacks using attack trees. IFIP ınternational conference on digital forensics. Springer, New York, NY, pp 331–343 96. Rekhis S, Boudriga N (2011) Logic-based approach for digital forensic investigation in communication Networks. Comput Secur 30(6–7):376–396

IOT Based Solution for Effective Social Distancing and Contact Tracing for COVID-19 Prevention S. Kanakaprabha, P. Arulprakash, V. Priyanka, Vineetha Varghese, and A. Sureshkumar

Abstract Coronavirus has infected billions of individuals worldwide, with the number of persons infected continuing to rise. Humans contract the virus through direct, indirect, or close contact with infected individuals. This proposed work introduces a new feature, an intelligent community distance system, that allows people to maintain community distances among others in the indoor and outdoor areas, to avoid exposure to COVID-19 and to delay its spread locally and internationally, to help prevent the spread of COVID-19. The proposed research intends to monitor an IoT-based portable monitoring device that is designed to measure COVID-19 signals. Furthermore, by monitoring real-time GPS data, the system automatically notifies medical authorities concerned about any confinement violations of patients who may be infected. Also, figure out what new tool will be beneficial for tracking and predicting COVID-19 collections. To support in the analysis of COVID-19, the solution incorporates a mobile system coupled with a portable device that is equipped with clever IoT capabilities (complex data analysis and intelligent data detection) embedded within the system. A comparison of various machine learning classifier algorithms such as SVM, Random Forest, KNN, and Decision Tree is presented as S. Kanakaprabha (B) · S. Kanakaprabha (B) · P. Arulprakash · V. Priyanka · V. Varghese · A. Sureshkumar Department of Computer Science and Engineering, Rathinam Technical Campus, Anna University, Coimbatore, India e-mail: [email protected] P. Arulprakash e-mail: [email protected] V. Priyanka e-mail: [email protected] V. Varghese e-mail: [email protected] A. Sureshkumar e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_46

629

630

S. Kanakaprabha et al.

the best model for making predictions and determining accuracy. We observed that KNN performs better, with a 95% accuracy rate. COVID-19 will be used to prevent the spread of diseases in future global medical problems using an automatic social distance monitoring and contact tracking system. Keywords Coronavirus · Decision tree · Global positioning system · Internet of things · Support vector machine

1 Introduction COVID-19 is a respiratory disease that is very infectious (SARS-CoV-2), which is very dangerous infectious virus. The respiratory system is attacked by SARS-CoV2, and the flu virus, which causes symptoms such as poisoning, fever, nausea and shortness of breath. To prevent the epidemic, many countries have it did Lockdown when the government forced it for residents to stay home during this critical time. Community health facilities, such as the Centers for Disease Control and Prevention (CDC), had to make it clear that avoiding close contact with other persons is the most efficient strategy to prevent the transmission of Covid-19. Surrounding dwellers of the earth have become accustomed to physical work in order to flatten the curves to the Covid-19 outbreak. Implementing community activities, team activities as well congregations like travel, meetings, gatherings, Workshops, prayers were banned during incarceration time. People are encouraged to use telephone and email manage and run events as much as possible to minimize personal contact. Continuing to contain the spread of virus, Adults are also recommended to follow good hygiene habits such as hand washing, wearing masks, and avoiding close contact with sick people. There is, however, a distinction between knowing what to do to minimize virus infection and truly doing it [1]. Using public broadcasts among persons at least one meter apart is one well-known method of avoiding the spread of Covid19. The coronavirus is spread from person to person through tiny droplets from the nose and mouth, according to WHO. To put it another way, social dispersal is the greatest strategy for limiting physical contact with probable corona-virus carriers by keeping at least a meter between people. The proposed work is to support Covid-19 antiretroviral measures. It provides a solution for finding people who gather in public places such as shopping malls, banks, market, temples, mosques, public transport and government offices etc. [2]. Monitoring and support of social isolation has evolved important in reducing the risk of internal transmission of disease crowded indoor areas [3]. Figure 1 shows in public areas, social precautions such as contrasting signs on the seats deter individuals from sitting next to each other. Social isolation and face mask is the best form of protection used in modern times. The rules set by the World Health Organization (WHO) provide a detailed review of external activities that one must strictly follow: WHO follow-up rules: (1) The face mask/cover must be worn before exposure. (2) A person leaving his or her private area must ensure that public

IOT Based Solution for Effective Social Distancing and Contact Tracing …

631

Fig. 1 Social distancing [4]

distance is maintained during his or her time in public places. As a result, there is no permanent solution in the absence of medical assistance. Few efforts are being made around the world to use IOT statistics as a preventative or predictable method against Covid-19, or as models in aspects of epidemiology. With the use of a food security camera that can analyses live or offline to detect social divisions and face mask coverage in the official, public, or public area, a vision tool for controlling social distance and face mask at work has been developed, will aid in the monitoring of health policies The Practicing Health protocol helps living longer and control rashes [5]. The spread of the COVID-19 epidemic has had a devastating effect on human mobility patterns as social behavior related to daily mobility. There is a need to understand the patterns of disease spread and its mechanisms among neighboring people in order to implement corrective measures at this time. To enhance the effectiveness of social media tracking, countries around the world are advancing advances in mobile technology and Internet of Things (IoT) to help keep track of common contacts to track people close to identified patients with COVID-19. Even with the advent of vaccine treatment the COVID-19 management approach will be continued until 2021, look to the foreseeable future in terms of digital communication that is an integral part of the response and use of preventive measures such as social isolation, and the use of masks. After several months of use of digital contact tracking technology, in-depth details of the suitability of the various methods and the usability, privacy, and trade-offs of ethical principles involved. In this propose, we provide a comprehensive analysis of digital communication tracking solutions in terms of their processes and technologies due to new emerging data on international information for the distribution of digital contact tracking technology. Contact tracking apps should establish data collection and data interpretation. Figure 2 shows contact tracing areas in public places.

2 Related Work Sengupta et al. [6] propose the need to develop a plan to respond to environmental outbreaks to assist in tracking and tracking safety-related concerns in industrial

632

S. Kanakaprabha et al.

Fig. 2 Contact tracing areas in public places

and community setting. Controlling infectious diseases and their spread necessitates complete communication. To track human activity, the framework will incorporate video feeds from surveillance cameras and IoT edge devices located in industrial or public spaces. The architecture suggested here is a hybrid method to integrating feeds from existing cameras and IoT devices with cloud-based computer-based edges. S. Srinivasan et al. suggests a complete and effective solution detection of a person, detection of a violation of social distance, detection of a face-to-face mask separation using object detection, integration and Convolution Neural Network (CNN) based on the binary category. In this case, YOLOv3, a local collection based on congestion audio applications (DBSCAN), Dual Shot Face Detector (DSFD) and MobileNetV2 based Binary classifier used for monitoring video databases. They done by comparative research for different facial expressions and facial mask classifications models. Finally, a video database labelling and labelling method is suggested Video data set to compensate for public data shortages and is used system testing. System performance is accurately assessed, Fl score as and predictive timing, which should be low enough to be applied in a practical way [7]. Khanfor et al. [8] to provide a comprehensive framework for improving pedestrian safety while roaming the real-world map of the smart city using the notion of Social IoT (SIoT).The purpose is reducing the risk of infection in very densely populated areas where social distance may not be properly addressed. Proposed the walkway recommends a pedestrian route in real time method while processing the movement of other devices. First, IoT devices are grouped into communities based on two SIoT relationships that took into consideration device locations as well as friendship norms among their owners. As a result, the weights on the city map roadways represented

IOT Based Solution for Effective Social Distancing and Contact Tracing …

633

their safety standards. Afterwards, they utilize the Dijkstra algorithm, which is a navigation algorithm, to recommend the safest route to go. Imitation effects were applied in the real-world IoT data set demonstrates this capability a proposed approach to achieving a trade between the two most secure and shorter routes depending on pedestrian preferences. This paper develops [9] a signal processing framework that enables for integrated topic movement analysis as well as automatic temperature testing. The system includes infrared-based sensors that use temperature data to track subject mobility and health. The sensors are connected to the network by existing IoT wireless devices planted according to different structures. The goal of this project is to link the local action of the headers by tracing their equal distance and route of arrival, as well as the remarkable discovery of body temperature in subjects near the IR sensors. Focus by the Bayesian methods, this paper also discusses good practices as well appropriate application implementation using field standards. For privacy neutrality, the proposed framework can be applied to it public and private services for health care, intelligent living and sharing space conditions without concern for privacy. W. Lv et al. propose an extended and unauthorized blockchain protocol, called by chain. Specifically, (1) the SRC confidentiality of the SRC protocol and the corresponding block structure are upgraded, by linking the anonymous evidence-based protocol with the key security mechanism. As a result, the connection between personal identity and on-chain location information is severed. At that point, the owner of the on-chain property may still claim ownership without disclosing the private key to anyone else. (2) Proposed field-based practice of incentives to propose to encourage IoT witnesses to advance the oversight of the monitoring industry. The suggested communication tracking and location verification technique works effectively in the actual world, according to several results. The suggested contract tracking protocol’s power consumption, time delay for each procedure, and BLE performance have all been studied to ensure the availability of tracking of digital communications in the actual world [10]. S. Arun Kumar et al. study proposes the concept of a smart wrist band with a heat sensor and IoT technology as a preventative approach. Blood pressure measures are also taken from time to time with the use of a blood pressure sensor. As a result, once the temperature or blood pressure are determined to be abnormal, this device helps to generate an alarm. With the use of IoT technology, faster information is transmitted on to the basic level user and second level relatives. As a result, by monitoring and notifying victims, this joint and active wrist band plays a critical role in saving lives. Because bacterial infections are linked to a rise in body temperature, our device will be extremely useful in detecting them early. Traditional measurement methods frequently necessitate human participation and are not of combined size. These issues are addressed in this model, which uses minimal control and sensor settings to handle temperature and blood pressure measurement, data processing, and storage [11]. Chloros and Ringas [12] the goal of this article is to look at the constraints and opportunities that applications that track transmission face, as well as how IoT systems might be used to record symptoms. The benefits and need for these applications’ development will be highlighted by evaluating their potential. The Fluspot

634

S. Kanakaprabha et al.

application was created specifically for this study. By boosting public awareness and providing timely information, Fluspot hopes to help prevent the spread of infections this season. Fluspot uses a wearable IoT device to carefully monitor flu flows and collect user inputs for viral propagation to the site. This anonymised and aggregated data is displayed on a map to provide a more accurate picture of the situation in each location. Another key element is that the artefact’s ability to monitor wearable signals is critical for users when it comes to using it in their daily lives. Waheed and Shafi [13] the study examines a number of technologies that are employed in a range of situations, including social isolation and prevention, isolation and isolation, COVID detection and evaluation, therapy and patient care, and hospital management. This study discusses transparent planning, technological techniques, and digital procedures, as well as the most up-to-date intelligent technologies in a range of disciplines that can aid in overcoming coronavirus intensity. IoT, AI, and machine learning play a significant role in the fight against COVID-19.AI has contributed significantly to the epidemic of resource management, public awareness, management of security measures, and assisting professionals in enforcing strict rules. Shubina et al. tracking of wearable contact is gaining more attention in the COVID-19 era in order to effectively prevent disease. Therefore, it is timely to identify available solutions for tracking wireless communication and their wearability. Existing contact monitoring app trading necessitates a detailed examination of technical skills such as accuracy, power consumption, availability, error sources when dealing with wireless channels, privacy concerns, and hurdles to larger apparel market access. We find, based on considerable literature study, that demarcated buildings, when compared to intermediate techniques, provide a better place to trade in terms of precision and user desire to use them, taking privacy considerations into account. This paper provides a brief overview of the technical solutions available for human tracking services, outlines key principles that affect the effectiveness of digital communication tracking, and presents a discussion of the potential impact of wear on coping with the spread of viral infections [14]. Luo et al. [15] proposes a model for the spread of infectious, contagious, infectious, asymptomatic disease, Diagnosis, Death (SEINRHD). The model was created using epidemiological data from COVID-19 in China and the estimation of social network heterogeneity. The original Wuhan public epidemic was recreated and updated with accurate data. We used this model to look into ways to manage the outbreak in instances when three-dimensional signals were not visible. The impact of undetectable cases on the spread of the epidemic was assessed based on effective replication rates, incidence of unusually high infections, and the type and structure of transmission. Management of asymptomatic cases can help reduce the curve of infection. Tracking of 75% of non-symptomatic cases is associated with a total reduction of 32.5% in new cases compared to asymptomatic and non-symptomatic tracking). The control and prevention of disease in families should be emphasized during the epidemic.

IOT Based Solution for Effective Social Distancing and Contact Tracing …

635

Sungheetha [16] the ongoing COVID-19 pandemic, there is a lot of interest in using smartphone-based Digital Proximity Tracing Technology (DPTT) for mitigation, containment, and monitoring due to the technology’s effectiveness and population acceptance. This research compares the Data-Driven Epidemic Intelligence Strategies (DDEIS) and DPTTs to highlight their differences and provide a fresh approach to address them. The essential components of DDEIS may be included to offer a social as well as technological solution for reducing the risk of epidemic revival, ensuring public health safety, and hastening the return of cities to normalcy. While evaluating its drawbacks and advantages for both societal decision-making and private decision-making, human behaviour is taken into account.

3 Proposed Method The proposed system consists of the following sub-systems: (1) Arduino Uno-based temperature measurement system (2) IoT system that measures the Arduino board via social distancing (3) on the server (4) A security guard smartphone application. To begin, everyone attempting to enter the residence must pass an unmodified temperature check. We’re utilising an Arduino Uno with an infrared thermometer (e.g., MLX906148) or a hot camera sensor for this (AMG88339 for example). It also employs the ESP8266 Wi-Fi module for MQTT protocol connection with Edge servers. If that person’s body temperature is abnormal, the door is locked, and a MQTT message with the temperature and the place where it was recorded is sent to the server. This message is received by the server, which then transfers it and carries on. with a security guard’s smartphone app, so they can arrive and make sure the person isn’t attempting to enter a work zone. In specific regions, Arduino board devices verify whether public distances are being utilised properly or not. Similarly, when public distances do not work well in particular rooms, a MQTT message will be sent to alert security personnel. The MQTT broker and the triple semantic store are used on the server side to provide message processing, event logging, reflection, and message transfer. Edge servers receive communications, do semantic annotations, and make assumptions to determine the appropriate security guard to notify. Security personnel use a basic Android smartphone app. Figure 3 is representing social distancing using smart device for Covid-19 is a programme that strives to guarantee that COVID-19 safety rules are followed correctly indoors. The proposed calculation will successfully give an answer for physical separating utilizing the ultrasonic sensor. The pseudo-code of the proposed calculation is referenced as follows (Fig. 4).

636

S. Kanakaprabha et al.

Fig. 3 Overall architecture of social distancing

Sensor System

IR Sensor

Ultrasonic Sensor

ML Algorithm

Mobile Alert

Fig. 4 Social distancing using smart device for COVID-19

3.1 IR Sensor There are two sorts of temperature measurement tools: touch and non-touch. Thermocouples, heat-resistant heat exchangers (RTDs), thermistors, and semiconductor temperature sensors are examples of infrared temperature sensors used in communication equipment. Because contact lenses measure temperature, they require physical contact with the object being measured to bring the sensor body up to temperature. When a relatively large sensor meets a small object and functions as a heat sink, the temperature of the object can be altered. Figure 5 IR sensor respectively.

IOT Based Solution for Effective Social Distancing and Contact Tracing …

637

Fig. 5 IR sensor

3.2 Ultrasonic Sensor Ultrasonic sensors use soundwaves to measure distance. The sensor head sends out an ultrasonic wave that is reflected back to it from the direction. Ultrasonic sensors use the time between output and reception to calculate the distance to the target. The ultrasonic sensor can help us to identify that are far away from the robot. The ultrasonic sensor, unlike the touch sensor, is not affected by physical contact. The range gives you plenty of room to react. For distances of 10 in. or more, an ultrasonic sensor is typically used, while for shorter distances, a light sensor is typically employed. ultrasonic sensors detect moving objects, and measure the relative position and movement of each object. The movement in the measuring area of each ultrasonic sensor is measured using a modified distance data conversion, and the vertical movement is measured using a measurement of the measurement range. Figure 6 showing ultrasonic sensor.

4 Social Distancing Algorithm It is the second step of our framework proposal. The suggested algorithm for measuring social distance serves two purposes. Function 1 aids in the identification of things in an image. It uses a detection method and provides human locations in the form of aggregate values such as XA (left), YA (top), XB (right), and YB (bottom). X = (XA + XB )/2

(1)

Y = (YA + YB )/2

(2)

638

S. Kanakaprabha et al.

Fig. 6 Ultrasonic sensor

where XA, YA, XB, and YB are compound numbers (left, top, right, bottom) of an item. X and Y are coin or centroid values. In addition, these parameters are transferred to the next function to measure social distance. Equation 3 shows the distance between the two items using the Euclidean distance, which establishes how close they are. When comparing this distance vector to the previously indicated threshold value, the decision was taken. If the Euclidean range is below a certain limit, it would be assumed that the two elements did not comply with the terms of social reduction or did not create sufficient distance between them.  (3) D = (X 2 − X 1)2 + (Y 2 − Y 1)2 where (X1, X2) and (Y1, Y2) are Centroid values of two objects. Pseudocode To identify human/object in the scope of physical removing. 1. 2. 3. 4. 5.

Input: Human/object in a scope of ultrasonic sensors. output: Alerting a sound sign to the client. Faculties the presence of people utilizing the ULTRASONIC sensor module. On the off chance that a human/moving item is identified, check for the distance. On the off chance that the distance is under 2 m, alert the sound directive for the distance in particular. 6. In the event that the distance is equivalent to 1.5 m, ready message and caution for safe separating and disinfection. 7. The message will be gone on till the individual isn’t cleaning. After the disinfection, individual needs to press the reset button.

IOT Based Solution for Effective Social Distancing and Contact Tracing …

639

4.1 Algorithm: Social Distancing Measurement

Input: IN: Image I containing N Number of frames of size 225x225x3 Output :D: Distance between two objects Initialize Parameter: Threshold = 90.0, //Distance between two objects Human_Count = 0; // Number of persons present within input scene default value = 0 Cvo = [], // Coordinate value of object Lcvo = [], // List of Coordinate value of objects Cen = [], // Centroid value of object Center = [], // List of Centroid values of objects Function Object_Coordinates(V) Picks = Human_Detection_Framework (VN) // provides the number of objects with their coordinate values. For (XA,YA,XB,YB) in Pick : Cvo = [XA,YA,XB,YB] C1 = ((XA + XB) * 0.5) //Centroid Value for Coordinate X // Centroid Value for Coordinate Y C2 = ((YA+YB)*0.5) Cen = [C1,C2] Center.append(Cen) Lcvo.append(Cvo) Human_Count +=1 End For Return Human_Count, Lcvo,center,Image End Function

4.2 Temperature Checking Using a contactless IR sensor, the temperature-checking device based on Arduino Uno detects the passenger’s temperature. One by one, the passengers go. If the temperature of the passenger is higher than the average human body temperature (37 °C), the Arduino Uno generates a signal to lock the door, preventing the person from entering the building, and sends a MQTT message indicating that a person with a high body temperature has been detected at a specific location. Otherwise, the door is opened to welcome the visitor inside.

640

S. Kanakaprabha et al.

5 Result and Analysis The IoT-based portable monitoring device is meant to measure the signals associated with COVID-19 and uses a machine learning model to anticipate the various machine learning techniques. Machine Learning techniques to increase the model’s accuracy and impact, as well as to avoid disease transmission in future global health issues. Figure 7 showing the various machine Learning algorithms so, With a 95% accuracy, KNN offers the best combination of performance values. The performance of two persons was examined in the remote sensing test, with the number of people within the distance view expected to decrease as the number of people increases. The performance of the distance monitor varies with the distance of objects from the camera, as it changes with respect to the measurement originally calculated between pixels and meters. The above graph shows an SVM accuracy is 60%, the Random forest accuracy is 70%, KNN accuracy is 98% and the decision tree accuracy is 50 respectively. Figure 8 shows the K-Means algorithms are used to calculate social distance. It is utilised to execute two points on the folks who have been detected. Because social distancing is tested between a minimum of two people, the cluster’s minimum necessary points are set to two, and the two-person distance parameter is set to two metres. The orange colour represents the safe people and the blue colour reprsents the unsafe peoples. If the space between two people is very small, it is considered risky; if the distance is greater, it is considered safe. Table 1 shows all of the models’ performance was evaluated using measures such as accuracy, specificity, precision, recall/sensitivity, and F1 score. The SVM is shows the accuracy rate is 60%, the precision rate is 63%, the specificity is 70%, the recall rate is 90%, and the F1-Score is 75%. Random Forest is shows the accuracy rate is 88%, the precision rate is 73%, the specificity is 75%, the recall rate is 75%, and the F1-Score is 90%. And the KNN is accuracy rate is 95%, the precision rate is 90%, Fig. 7 Machine learnings algorithms

IOT Based Solution for Effective Social Distancing and Contact Tracing …

641

Fig. 8 Social distance measurement

Table 1 Classification result for four models Models

Evaluation metrices TP

TN

FP

FN

Accuracy

Specificity

Precision

Recall

F1 score

SVM

28

6

2

3

0.60

0.70

0.63

0.90

0.75

Random forest

29

6

2

2

0.88

0.75

0.73

0.75

0.90

KNN

29

5

3

2

0.95

0.89

0.90

0.93

0.92

Decision tree

26

6

2

5

0.65

0.75

0.69

0.73

0.72

the specificity is 89%, the recall rate is 93% and the F1-Score is 92%. Decision tree is shows the accuracy rate is 65%, the precision rate is 69%, the specificity is 75%, the recall rate is 73% and the F1-Score is 72%. The accuracy value of four models reveals that KNN is reliable for monitoring IoT-based portable monitoring device is developed to measure COVID-19 signals. With the highest F1 score of the four models, KNN emerges as the best. Although the SVM model is relatively similar to the decision tree, it cannot be regarded a robust model because to its low recall of 0.83.

6 Conclusions The proposed tool is based on real-time sensors such as infrared (IR) and ultrasonic sensors for effective social distancing and contact tracing for COVID-19 prevention. Temperature, heart rate, SpO2 , and cough rate are all measured using the wearable

642

S. Kanakaprabha et al.

sensor layer. It also sends real-time patient GPS position data to medical administrators and notifies family members to alleviate stress. The peripheral interface of the app is responsible for storing, collecting and analysing data in order to monitor and control public life and administration during the epidemic. The Android mobile app is extremely helpful in informing family members about patient status and lowering transmission rates. The wearable device has been fully designed to monitor patient health during and after infection. In order to control, monitor, and control patients who may be infected with COVID-19 in the spread of the disease, this system was tested and validated in real time at a hospital. A wearable device might be used as a model, allowing airport passengers to sit alone as they enter and exit. This work has undergone significant investigation in order to deliver the greatest device performance by comparing existing domains. The project’s new capabilities are useful for evaluating health symptoms, tracking and monitoring a patient during detention, storing data to predict the scenario, and contacting authorities in a timely manner so that they may be properly monitored and use the Android platform to stay informed. The patient status of family respondents. Our proposed tool can also be used to prevent the spread of disease in future global health problems. And also take the advantage of this proposed technology that can help to identifying the early symptoms and treatment.

References 1. Hou YC, Baharuddin MZ, Yussof S, Dzulkifly S (2020) Social distancing detection with deep learning model. In: 2020 8th international conference on information technology and multimedia (ICIMU), 2020, pp 334–338. https://doi.org/10.1109/ICIMU49871.2020.9243478 2. Ahamad AH, Zaini N, Latip MFA (2020) Person detection for social distancing and safety violation alert based on segmented ROI. In: 2020 10th IEEE international conference on control system, computing and engineering (ICCSCE), 2020, pp 113–118. https://doi.org/10.1109/ICC SCE50387.2020.9204934 3. Motlagh NH et al (2021) Monitoring social distancing in smart spaces using infrastructurebased sensors. In: 2021 IEEE 7th world forum on internet of things (WF-IoT), 2021, pp 124– 129. https://doi.org/10.1109/WF-IoT51360.2021.9595897 4. https://binged.it/3rD1btm 5. Sathyabama B, Devpura A, Maroti M, Rajput RS (2020) Monitoring pandemic precautionary protocols using real-time surveillance and artificial intelligence. In: 2020 3rd international conference on ıntelligent sustainable systems (ICISS), 2020, pp 1036–1041. https://doi.org/10. 1109/ICISS49785.2020.9315934 6. Sengupta K, Srivastava PR (2022) HRNET: Ai-on-edge for mask detection and social distancing calculation. SN Comput Sci 3:157. https://doi.org/10.1007/s42979-022-01023-1 7. Srinivasan S, Rujula Singh R, Biradar RR, Revathi S (2021) COVID-19 monitoring system using social distancing and face mask detection on surveillance video datasets. In: 2021 international conference on emerging smart computing and ınformatics (ESCI), 2021, pp 449–455. https://doi.org/10.1109/ESCI50559.2021.9396783 8. Khanfor A, Friji H, Ghazzai H, Massoud Y (2020) A social IoT-driven pedestrian routing approach during epidemic time. In: 2020 IEEE global conference on artificial intelligence and ınternet of things (GCAIoT), 2020, pp 1–6. https://doi.org/10.1109/GCAIoT51063.2020.934 5900

IOT Based Solution for Effective Social Distancing and Contact Tracing …

643

9. Savazzi S, Rampa V, Costa L, Kianoush S, Tolochenko D (2021) Processing of bodyinduced thermal signatures for physical distancing and temperature screening. IEEE Sens J 21(13):14168–14179. https://doi.org/10.1109/JSEN.2020.3047143 10. Lv W, Wu S, Jiang C, Cui Y, Qiu X, Zhang Y (2022) Towards large-scale and privacy-preserving contact tracing in COVID-19 pandemic: a blockchain perspective. IEEE Trans Netw Sci Eng 9(1):282–298. https://doi.org/10.1109/TNSE.2020.3030925 11. Arunkumar S, Mohana Sundaram N, Ishvarya D (2021) Temperature sensing wrist band for Covid-19 crisis. In: 2021 international conference on advancements in electrical, electronics, communication, computing and automation (ICAECA), 2021, pp 1–5, https://doi.org/10.1109/ ICAECA52838.2021.9675689 12. Chloros D, Ringas D (2020) Fluspot: seasonal flu tracking app exploiting wearable IoT device for symptoms monitoring. In: 2020 5th south-east Europe design automation, computer engineering, computer networks and social media conference (SEEDA-CECNSM), 2020, pp 1–7. https://doi.org/10.1109/SEEDA-CECNSM49515.2020.9221843 13. Waheed A, Shafi J (2020) Successful role of smart technology to combat COVID-19. In: 2020 fourth international conference on I-SMAC (IoT in social, mobile, analytics and cloud) (I-SMAC), 2020, pp 772–777. https://doi.org/10.1109/I-SMAC49090.2020.9243444 14. Shubina V, Ometov A, Simona Lohan E (2020) Technical perspectives of contact-tracing applications on wearables for COVID-19 control. In: 2020 12th ınternational congress on ultra modern telecommunications and control systems and workshops (ICUMT), 2020, pp 229–235. https://doi.org/10.1109/ICUMT51630.2020.9222246 15. Luo T, Cao Z, Wang Y, Zeng D, Zhang Q (2021) Role of asymptomatic COVID-19 cases in viral transmission: findings from a hierarchical community contact network model. IEEE Trans Autom Sci Eng. https://doi.org/10.1109/TASE.2021.3106782 16. Sungheetha A (2021) COVID-19 risk minimization decision making strategy using data-driven model. J Inf Technol 3(01):57–66

Design and Implementation of Highly Secured Nano AES Cryptographic Algorithm for Internet of Things E. Roopa and Yasha Jyothi M. Shirur

Abstract Advancement in the internet of things to meet the requirement of human beings and society makes integration of multiple devices into a single system. The integration of hardware and software needs to be provided with security to avoid the stealing of the data. Otherwise, the hacker may gain control over the devices and change the functioning of the system which may lead to malfunction. In order to provide security for the data transfer in IoT, the security algorithm need to be embedded with the IoT. The algorithm should provide high security and at the same time, it should be efficient. In this paper, an attempt is made to design a synthesizable Deoxyribonucleic acid (DNA) based Nano Advanced Encryption Standard (AES) Intellectual Property (IP) Core which can be used as a crypto engine in an IoT system. The crypto engine developed is optimized in terms of power, area and delay. The developed design when compared with the conventional design has given an area advantage of 81.6%, power of 21.17%, gate delay of 88.44% and path delay of 99.64%. Keywords Internet of things (IoT) · Hacker · Crypto engine · Deoxyribonucleic acid (DNA) · Nano · Advanced encryption standard (AES) · Intellectual property (IP) core

E. Roopa (B) Department of ECE, VLSI Design and Embedded Systems, BNM Institute of Technology, Bengaluru, Karnataka, India e-mail: [email protected] Y. J. M. Shirur Department of ECE, BNM Institute of Technology, Bengaluru, Karnataka, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_47

645

646

E. Roopa and Y. J. M. Shirur

1 Introduction 1.1 Internet-of-Things IoT is a physical object or network that is inbuilt with software, sensors and smart techniques and algorithm for data exchange. The data exchange will happen between two devices or multiple devices through the internet and it’s termed as machine-tomachine (M2M) Communication. Every year the no of devices connecting to IOT is increasing rapidly, based on the statistics given by IoT Analytics, by 2025 the global number of connected IoT devices will be around 21.5 billion which is a 17% growth increased when compared to 2018 [1]. The devices include only active devices/nodes or gateways that interact with the end point sensors. It is because of smart techniques and algorithms embedded with it which supports wired and wireless technologies. The main motto behind integration is proper connectivity, low power consumption, low-cost sensor technology and to provide high security for data transfer. It can be accomplished by wireless communication and sensor networks, VLSI Design, artificial intelligence, machine learning, and cloud computing analytics. Figure 1 shows sudden growth in IoT devices with the technology supporting [1]. Nowadays, the whole world is full of IoT devices and providing security is the major concern while developing the IoT systems for applications like military, agriculture, educational system and even in health care where the data transfer is confidential and to be transferred securely. There may be multiple attacks which need to be addressed. Some of the major attacks usually dealt with are stealing information, taking control over the system and disrupting services [2].

Fig. 1 Statistics to show rapid growth of IoT devices [1]

Design and Implementation of Highly Secured Nano AES …

647

Fig. 2 Attacks addressed in IoT system to provide security [2]

The solution for the above attack is secured data transfer, which is done by embedding the crypto engine into the IoT system. Figure 2 shows the major attacks which are considered to be addressed before the IoT systems are rolled out commercially for usage by the mankind.

1.2 Crypto System Crypto System involves encryption and decryption. In the encryption process the plain text will be changed into an unreadable format, called cipher text. The cipher text will be changed into plain text in the decryption process. Figure 3 shows the block diagram of traditional crypto system. The classification of cryptography algorithms is shown in Fig. 4. It is basically classified as symmetric and asymmetric key cryptography, further the classification

Fig. 3 Traditional crypto system

648

E. Roopa and Y. J. M. Shirur

Fig. 4 Classification of cryptography algorithms [3]

is made based on the cipher text. In order to suit the requirement of the applications, there are wide ranges of cryptography algorithms are there in the literature [3]. DES—Data Encryption Standard RC—Rivest Cipher RSA—Rivest, Shamir, and Adelman DSA—Digital Signature Algorithm. They use different approaches to transfer the data securely. Some algorithms have drawbacks to transfer the data through a secured network. Since 2001 Advanced Encryption Standard (AES) is the highly secured algorithm used for data transfer, it over comes all the draw backs of other algorithms. It performs 10 rounds of operation for 128-bit input along with a 128-bit key and provides 128-bit output. This is the standard AES algorithm by the inventor Rijndael. This process is called as Rijndael encryption. The highlight of this algorithm is providing different keys in each round with the key generation technique, so that the secured transfer of data is possible. It is impossible to retrieve the data by third parties. In this work, the AES is implemented based on the concept of complex DNA structure to provide efficient and high security for data transfer in IoT connected devices.

1.3 Advanced Encryption Standard (AES) Advanced Encryption Standard (AES) is a type of block cipher, which uses an encryption (secret key) key and many rounds of operation. This block cipher at a time uses

Design and Implementation of Highly Secured Nano AES …

649

a single block of data [4]. AES algorithm is a mathematical description, that reveals the procedure to encrypt data. It is not a source code or computer program to do the process on its own.

1.4 Deoxyribonucleic Acid (DNA) DNA is an essential component in living organisms. The functioning of all living organisms depends on DNA [5]. This is basically used to store genetic information. This information is unique, it can’t be copied or duplicated. The idea behind biotechnology is that, it is not possible to change or steal the information while transmitting from sender to receiver. That is why it is called as DNA cryptography. Table 1 gives the details of DNA encoder symbols. It uses 40 characters out of which 26 (A–Z) alphabets, 10 (0–9) numbers and 4 special characters (space, comma, full stop and semicolon). DNA Coding Technology Table 2 clearly explains how the conversion will be performed. In this technique, each alphabet is given by a set of three letters. Every alphabet has its own notation. This three-group set is converted into decimal code. This decimal code is changed and is represented into an 8-bit binary code. It can be observed the input data is a character which is finally converted into a triplet code after subjecting it to a processing unit. Table 1 DNA encoder symbols [6] Character

DNA symbol

Character

DNA symbol

Character

DNA symbol

Character

DNA symbol

A

CGA

K

AAG

U

CTG

0

ACT

B

CCA

L

TGC

V

CCT

1

ACC

C

GTT

M

TCC

W

CCG

2

TAG

D

TTG

N

TCT

X

CTA

3

GCA

E

GGC

O

GGA

Y

AAA

4

GAG

F

GGT

P

GTG

Z

CTT

5

AGA

G

TTT

Q

AAC

Space

ATA

6

TTA

H

CGC

R

TCA

Comma (,)

GAT

7

ACA

I

ATG

S

ACG

Full stop (.)

GAT

8

AGG

J

AGT

T

TTC

Semicolon (;)

GCT

9

GCG

650

E. Roopa and Y. J. M. Shirur

Table 2 ASCII codes Alphabet

DNA conversion

Decimal code

Binary code

A

CGA

67 71 65

01,000,011 01,000,111 01,000,001

B

CCA

67 67 65

01,000,011 01,000,011 01,000,001

C

GTT

71 84 84

01,000,111 01,010,100 01,010,100

D

TTG

84 84 71

01,010,100 01,010,100 01,000,111

E

GGC

71 71 67

01,000,111 01,000,111 01,000,011

F

GGT

71 71 84

01,000,111 01,000,111 01,010,100

0

ACT

65 67 84

01,000,001 01,000,011 01,010,100

1

ACG

65 67 71

01,000,001 01,000,011 01,000,111

2

TAG

84 65 71

01,010,100 01,000,001 01,000,111

3

GCA

71 67 65

01,000,111 01,000,011 01,000,001

4

GAG

71 65 71

01,000,111 01,000,001 01,000,111

2 Literature Survey The highlights of the literature survey are show in Table 3. Some of the Differential Power Analysis (DPA) attacks are vulnerable to Advanced Encryption Standard algorithms. It is giving an approximate value rather than exact values. For different applications, different approaches and methods are used. There is no exactly single methodology for all applications. Power consumption is reduced in some approaches but there is a trade-off with respect to resource usage. The available approaches make use of both combinational logic and sequential logic, and switching between these two consumes more power. Therefore, in the developed algorithm an attempt is made to replace some part of sequential logic to avoid the switching between the logic and in turn reduce noticeable power.

3 Design Implementation AES implementation consists of two main blocks, one is state-register and other one is key-register. In the implemented technique the state-register itself acts as a shiftRow. The implementation is performing 128-bit encryption [7]. The total process is performed in 10 cycles, in first compute—1 round per cycle, here hardware of each round is being reused to save the area against a fully unrolled implementation. The proposed AES design is shown in Fig. 5. The inputs plain text and key are given to the state-register and key-register, both are sharing the same s-box, and they are merged to reduce the area [8]. This is the main idea of the proposed method. The flow is as follows; SubBytes, ShiftRows, MixColumn and AddRoundKey [9]. It repeats until 9 rounds, then the final round operation will be performed. In the final round there is no MixColumn. The output is given to the DNA encoder, and it encodes the

Design and Implementation of Highly Secured Nano AES …

651

Table 3 Highlights of the literature survey SI. No.

Title of the paper

Authors

Publication details

Description

1

Differential power analysis: a serious threat for FPGA security

M. Masoumi

International Journal of Internet Technology and Secured Transactions 4(1):12–25

Using differential power analysis (DPA) is a good practice for attack measurement

2

The research of DPA attacks against AES implementations

Yu HAN, Xuecheng Zou Liu Zhenglin, and Yi-cheng CHEN

The Journal of China Universities of Posts and Telecommunications 15(4):101–106

Hardware implementation has less data-dependent power leakages to resist power attacks

3

AES against first and J. Zhou and second-order differential M. Yung, Eds power analysis applied cryptography and network security

Vol. 6123, Springer-Verlag, pp. 168–185. Berlin, Germany

AES is secure for software implementation to be resistant against first and second order DPA in practice

4

High-speed VLSI architectures for the AES algorithm

X. Zhang and K. K. Parhi

IEEE Transaction Very Large Scale Integrated (VLSI) System, vol. 12, no. 9, pp. 957–967

The speed will be increased using combinational logic

5

Post-quantum crypto processors optimized for edge and resource-constrained devices in IoT

Shahriar Ebrahimi, Siavash BayatSarmadi, Hatameh MosanaeiBoorani

IEEE Internet of Things Journal PP (99):1–1

Optimized variant for binary learning with errors over the ring (Ring-LWE) is used against quantum attacks

6

A high data rate pipelined architecture of AES encryption/decryption in storage area networks

Hossein Kouzehgar, Meisam Nesary Moghadam and Pooya Torkzadeh

26th Iranian Conference on Electrical Engineering (ICEE2018)

High throughput can be achieved through pipeline structure

7

A high throughput and secure authentication-encryption AES-CCM algorithm on asynchronous multicore processor

Ali Akbar Pammu, Weng- Geng Ho, Ne Kyaw Zwa Lwin, Kwen- Siong Chong and Bah-Hwee Gwee

IEEE Transactions on Information Forensics and Security PP (99)

High through can be achieved using parallel-encryption implemented on an asynchronous multicore processor (AMP-MP)

652

E. Roopa and Y. J. M. Shirur

Fig. 5 Block diagram of the proposed nano-AES crypto system

Fig. 6 Key generation process

data. The encoded data is transmitted over the allocated network. At the receiver side the encoded data is decoded using the DNA decoder, then it would be given to the decrypted to perform the decryption. In the key generation process, first data is Rot Word (Rotation Word), it takes out the last column of the data and rotates the column one time, then change the data with s-box. That data is XOR-ed with the first column of the actual data. This data is XOR-ed with an RCON, called Round Constant [8], which is shown in Fig. 6. So new keys will be generated for each and every round. This total process is explained in the flow chart, which is given in the Fig. 7. Flow Chart of AES Design See Fig. 7.

4 Result and Discussions Nano-AES crypto algorithm is coded in Verilog and simulated in ModelSim and synthesized in Xilinx9.1. Figure 8 depicts the simulation results obtained; the details

Design and Implementation of Highly Secured Nano AES …

653

Fig. 7 Execution process of AES design

of the marker representation are as follows: 1. Input Key 2. Plain Text 3. Encrypted Data 4. Decrypted Data 5. Encryption Acknowledgement 6. Decrypted Acknowledgement. It is to be observed that the plain text and the decrypted data are the same. The obtained results are cross verified using the online AES calculator. The Nano DNA AES encoder and decoder output simulation results are shown in the Fig. 9, the details of the marker representation are as follows: 1 Encrypted Data 2. DNA Encoder output 3. DNA Decoder output. Figure 10 shows the device utilization chart. The time taken by the encryption process is 8.801 ns, which is depicted in the synthesis report shown in Fig. 11. By using the new approaches in the proposed algorithm, the area, delay and power have reduced. This can be observed from the bar graphs shown in Figs. 12, 13 and 14. Table 4 shows the comparison between existing and proposed design with clock gating.

Fig. 8 Nano-DNA AES encryption and decryption with clock gating

654

E. Roopa and Y. J. M. Shirur

Fig. 9 Simulated results obtained for Nano DNA AES encoder and decoder

Fig. 10 Device utilization chart

Fig. 11 Synthesis timing report for encryption

Design and Implementation of Highly Secured Nano AES … Fig. 12 Comparison of area utilization

655

25000

20000

15000

10000

5000

0 LUT Normal AES Design

Fig. 13 Comparison of delay

Slices

Gates

Proposed Nano AES with clock Gating

180 160 140 120 100 80 60 40 20 0 Overall Delay Normal AES Design

Fig. 14 Comparison of power consumption

Gate Delay

Path Delay

Proposed Nano AES with clock Gating

656

E. Roopa and Y. J. M. Shirur

Table 4 Comparison between conventional AES design with proposed nano AES with clock gating Sl. No.

Method name

1

Conventional AES design

2

3

Area Slice

Flip flops

LUT

Max delay Gate (ns) delay (ns)

Path delay (ns)

Total power (mw)

7734

21,207

21,207

160.860

25.302

135.558

23,032

Proposed nano 1670 AES with clock gating

1066

3900

3.405

2.923

0.482

18,155

Percentage reduction (%)

94.9

81.6

78.4

Delay

97.88

88.44

99.64

21.17

5 Conclusion The light-weight AES architecture is the best choice for resource-constrained IoT devices. It uses a symmetric cryptography algorithm and provides high security. This algorithm is used by many applications and networks. The design contains two main registers each is of 8 bits, one is State-register and the next one is a key-register. It is designed in a way that the ShiftRows are running inside the state-register to reduce the logic. SubBytes are optimized for these two registers as they both uses the same S-Box. Next is a mix-Column with 8-bit input and output, with this low-area design is achieved. The area and power consumption are reduced by using the clock gating technique. The crypto engine developed is optimized in terms of power, area and delay. The developed design when compared with the conventional design has given an area advantage of 81.6%, power of 21.17%, gate delay of 88.44% and path delay of 99.64%. Acknowledgements Authors wish to thank the management of BNM Institute of Technology and Visvesvaraya Technological University for their support and encouragement rendered to in this work.

References 1. Lueth KL (2018) State of the IoT 2018: number of IoT devices now at 7B—market accelerating. IoT Analytics 2. Joe (2015) The 10 challenges of securing IoT communications 3. Elgeldawi E, Mahrous M, Sayed A (2019) A comparative analysis of symmetric algorithms in cloud computing. Int J Comput Appl 4. Biryukov A (2015) Block ciphers and stream ciphers: the state of the art 5. Condon A (2006) Designed DNA molecules: principles and applications of molecular nanotechnology. British Columbia V6T 1Z4, Canada 6. Miles I (2020) An in-depth guide to AES encryption with angular service implementation

Design and Implementation of Highly Secured Nano AES …

657

7. Arrag S, Hamdoun A, Tragha A, Khamlich S (2012) Design and implementation a different architectures of mixcolumn in FPGA. Research Gate 8. Waqas U, Afzal S, Mir MA, Yousaf M (2015) Generation of AES like S-boxes by replacing affine matrix. Research Gate 9. Oh J-Y, Yang D-l, Chon K-H (2010) A selective encryption algorithm based on aes for medical information. Healthc Inf Res

Convergence of Communication Technologies with Internet of Things V. Dankan Gowda, Suma Sira Jacob, Naziya Hussain, R. Chennappan, and D. T. Sakhare

Abstract Internet of Things (IoT) is the term used to describe a network of physical things such as mobile devices and household appliances that are embedded with electronics, software, sensors, and network connection that enables these objects to gather and exchange data. Sensors, recognition and remote control of items are all made possible by the Internet of Things (IoT). Once this property is combined with sensors and actuators, it becomes an example of a cyber-physical system, which includes technologies like intelligent power grids (grids), intelligent homes (smart homes), smart cities (smart cities), and intelligent transportation systems (ITS). Integrating MANET and WSN with IoT is covered in this study. Technology and protocols needed to deploy the Internet of Things (IoT) are explored in this article. Keywords Internet of things · Wireless sensor network · Protocol · Network · Sensor · Node

V. Dankan Gowda (B) Department of Electronics and Communication Engineering, BMS Institute of Technology and Mangement, Bangalore, Karnataka, India e-mail: [email protected] S. S. Jacob Department of Information Technology, Sri Krishna College of Technology, Coimbatore, Tamil Nadu, India N. Hussain School of Computers, IPS Academy, Indore, Madhya Pradesh, India R. Chennappan Department of Computer Science, Karpagam Academy of Higher Education, Coimbatore, Tamil Nadu, India D. T. Sakhare Department of Chemistry, U.G., P.G. and Research Centre, Shivaji Arts Commerce and Science College, Kannad Dist. Aurangabad, Maharashtra, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_48

659

660

V. Dankan Gowda et al.

1 Introduction Kevin Ashton was the first to use the phrase “Internet of Things” in relation to supply chain management. As a result, the concept has evolved over the last decade to include a wider range of applications, including healthcare, utilities, transportation, and more [1]. Computers can now discern information without any human intervention, even if the concept of ‘Things’ has evolved throughout time with advancements in technology. One of the most significant developments in the recent decade has been the advent of the Internet of Things concept. It is now possible to turn an isolated item into a communicative device because to advancements in the electronics industry, the expansion of communication protocols, and the shrinking size of devices and transceivers. As a result, compact processing and sensor devices have become more powerful, more energy efficient, and more capable of storing data. The exponential growth in the number of Internet-connected sensing and computing devices known as “smart gadgets” has been sparked by these advancements in electronics and computer science. These devices may deliver massive services only limited by human imagination. The Internet of Things (IoT) is a network of interconnected physical devices that communicate and exchange data with one another. Sensors, radio frequency identification devices (RFID), infrared sensors (IR), laser scanners, Global Positioning System (GPS), wireless LANs (WLAN), and even Local Area Networks (LANs) interfaces [2] are required for these devices. By using IPv6, these devices may be linked to the Internet and managed from a distance. Sensors may convey data to other devices for processing, allowing these devices to communicate with one other. An M2M connection is one in which a machine communicates with another machine under the IoT paradigm. The Internet of Things (IoT) relies heavily on WSN and RFID technology [3]. In recent years, a number of studies have been conducted to examine the increasing use of IoT applications. The integration of heterogeneous devices, including mobile phones, laptop systems, PDAs and smart phones, wireless sensors and network-enabled physical items, such as RFID and smart visual tags, is also taking place. Once they’ve been integrated, these gadgets will be able to communicate easily with the Internet [4]. When it comes to building smart cities, IoT will play a critical role in ensuring its residents have access to high-quality amenities and services. To reach this purpose, it is necessary to gather data from a variety of sources, analyse it, and display it in a variety of ways. This action requires a great deal of standardisation work from several angles. It is critical that new, self-governing and adaptable smart city services be developed that can be used in a variety of application areas, from environmental monitoring to security control and sophisticated applications. Monitoring of natural resources such as air, water, noise, and light pollution, as well as pollution caused by human activity such as automobiles, industries, and traffic, are all included in environmental monitoring. Monitoring natural resources, military operations and others are included in this category as well. Structural monitoring to avoid bridge and ancient building collapses, support for people living and travelling, assistance for elderly and handicapped persons, emergency response are all examples of security control. Smart TV, uninterrupted multimedia streaming,

Convergence of Communication Technologies with Internet of Things

661

online gaming through mobile devices, and a host of other high-tech applications are just a few examples. In smart cities, WSNs and MANETs are key technology for a variety of IoT applications [5]. These two technologies are more suited for the installation of IoT applications because of their localised and self-configuration capabilities. The implementation of WSNs and MANETs in cities for public safety, location-aware computation, and environmental monitoring is predicted by several studies to occur shortly [6]. Mobile phones equipped with several wireless interfaces (IEEE 802.11, Bluetooth, and 3G) have already begun delivering low-power connection alternatives, such as IEEE 802.15.4, onboard. It is projected that in the near future, most consumer gadgets will have access to low-power connection [7].

2 Literature Survey The Internet of Things (IoT) is a network of things that are linked through the Internet. Internet of Things (IoT) is an amalgamation of WSNs and the old Internet. Resourceconstrained sensors are the norm in WSNs, but powerful devices predominate over them on the internet. Consequently, IoT may be described as a network of diverse gadgets. There are many different types of Internet of Things devices, from sensors to light bulbs to energy metres to vehicle parts to smartphones to PCs and tablets to powerful servers to cloud computing systems. As a result, the Internet of Things has the potential to link billions of IP-enabled devices. In [8], authors conducted a survey on WSN assaults, and in Raymond and, authors reported their findings. WSN DoS attacks and responses were examined in [9] by the authors. As outlined in [10], IP-based WSN security concerns and responses were discussed in depth in this paper. Many studies have shown that there are still many unsolved problems. An in-depth look at the hurdles and issues linked with IoT is provided in [11] author’s work. To name just a few, there’s interoperability, an IoT-based business model that allows for the networking of billions of devices, security and privacy issues such as trustworthiness and complete end-to-end encryption, and more. Security models for IoT technologies must be efficient, according to the study authors. A different point of view was expressed in [12] by the authors, who pointed out that there are two primary concerns: security and privacy. They’ve gone through the obstacles and spoken about them from various angles. Data storage and secure processing have all been taken into consideration while addressing security concerns, including authentication and authorisation as well as Denial-of-Service (DoS) attack mitigation. Additionally, while discussing privacy issues, they explored the privacy choices, identity control, and commercial demands of passive users. Among the issues discussed in [13] are those relating to the security of the Internet of Things. The authors made it apparent that large-scale implementation of IoT would be fraught with difficulty and would need careful consideration of a number of significant issues. Some of the difficulties that need to be addressed include those related to secrecy and privacy, security, heterogeneous device management, and network restrictions. By [14], the security risks and difficulties of the Internet of Things are addressed in an innovative way [15].

662

V. Dankan Gowda et al.

As the most important goal of IoT security, data security is the basis for this categorization. Sleep Deprivation (SD) attacks have been studied in a variety of MANET setups. The Route Request (RREQ) flooding attack in MANETs was discussed in [16]. They came up with a way to avoid the RREQ flooding attack by relying on the monitoring of the neighbouring node. The incoming RREQs are kept in a priority queue. If RREQs are regularly produced by a single node, their priority is degraded. In a MANET environment, attackers may drain the power of wireless devices like PDAs and notepads in a variety of methods, as described in [17]. The battery life of notepads and PDAs was assessed in a series of trials done under this assault. Finally, they observed that this assault consumes the battery’s energy faster. In order to defeat this attack, they’ve created a power-secure architecture that makes use of system energy monitoring and layered authentication. AODV protocol flooding may be reduced by using the session-based history table provided in [18]. The average number of RREQ packets is logged and compared to the discard threshold in this method to identify flooding attacks. In order to conduct a sleep deprivation assault, adversary nodes must become cluster chiefs, as predicted in [19]. Random vote, round robin, and hash-based schemes are examined for minimising sleep deprivation attack mitigation [20]. When a node detects an incursion, the agent then begins a global reaction. Analyzing the behaviour of neighbouring nodes may help assess the trustworthiness of other nodes in a MANET, as stated by [21]. An investigation into the use of mobile agents in MANETs led to the conclusion that mobile agentbased IDS is the best option for MANET IDS. The Dendritic Cell Algorithm is a well-known addition to the danger project (DCA). It takes advantage of the HIS’s innate resistance to dendritic cells. The DCA method therefore proves that it can identify port scanning attacks, and it is thus argued that DCA may be utilised as an algorithm for anomaly detection. A comparison of MANET and sensor network environment features with innate immunity traits has been made by [12, 13]. This shows that the DCA may be used to identify various forms of assaults in dynamic settings like MANET.

3 RFID and IoT To make the Internet of Things work, a few basic procedures are required [8]. In order to implement M2M and D2D communication, two major needs must be met: The first step is to recognise what you’re looking at. In order to be uniquely identified, objects must have an integrated Auto-ID technology, which might be an RFID tag. Then there’s the issue of communication. Devices using RFID tags can access the most important information. The tags’ RFID readers may transmit the data they acquire to the internet. In Fig. 1, you can see how RFID technology may be linked to the Internet. Since the RFID readers are working as interpreters, they’re useful in this situation. Radio-frequency identification tags may be passive or active. The radio signal provided by the reader serves as the only source of power for passive tags, which need no external power source. As a result, passive tags are far less expensive

Convergence of Communication Technologies with Internet of Things

663

Fig. 1 RFID technology connected with the internet

than active tags. The Electronic Product Code is linked to RFID’s involvement in the Internet of Things (IoT) (EPC). The MIT Auto-ID Center has developed EPC to serve as a global identifier for all physical objects on the planet. Since RFID readers function as sensor nodes and RFID technology is the wireless connection used to gather data from tags, all the WSN techniques may be used to RFID communication. There is a server linked to the Internet through an RFID reader network.

4 Wireless Sensor Network The sink node in a wireless sensor network collects data from the network’s sensor nodes and distributes it to the rest of the network’s nodes [9]. Sensor nodes are often used to gather environmental data, such as temperature, pressure, humidity, and proximity, among other factors. There are two ways to look at WSN in relation to the Internet of Things: The network as a whole is a single entity, with the sink node serving as the primary access point for all network information. In terms of memory, compute, and battery power, WSNs are well-known to be constrained in their capabilities. Changing or recharging the small batteries that power WSN nodes is impractical in a WSN setup. To minimise frequent disconnections, nodes must save battery power by using efficient transmission techniques and implementing good MAC and routing protocols. In a monitored environment, WSN

664

V. Dankan Gowda et al.

may be installed using a star, mesh, or tree topology. Because the nodes in a star topology are all within one hop of the sink node, data gathered from several sensors may be redundant. Such duplicate data is processed by the sink node in this case. With the usage of a MANET communications overlay in both mesh and tree topologies, multi-hop communication may take place. One method of cutting down on the network’s power consumption is to use a graph theory-based algorithm, such as the minimal Connected Dominant Set (CDS) or minimum Spanning Tree. WSN has two significant variants: For the first time, a patient’s bodily status may be monitored and sent to the Internet through a wireless body area network (WBAN) including a number of tiny biosensors [5]. Using the WBAN for a smart health monitoring system is the primary use case. The Wireless Sensor and Actuator Network (WSAN) [7] is the second option. Sinks, sensors, and actuators are the three main kinds of nodes in WSANs. Data from sensors is sent to the sink node, which processes the information and sends it to the backbone server through a link. Actuators are able to respond to orders from the server and control the environment.

5 Connecting WSN with IoT Because sensor nodes are small and have limited resources, establishing a WSN connection to the Internet is very difficult. To connect WSNs to the Internet, many designs have been suggested. In the image, these designs may be divided into three classes [12]. Figure 2, IP over WSN, sensor over IP, and higher-level gateway overlays (i.e., the IP layer on top of the sensor layer). IP overlay over WSN: Sensor nodes are given a distinct IP address so that they may communicate with the Internet. Due to the low processing capability of the sensor nodes, this strategy is not a viable one. This problem may be alleviated in the future if efforts are made to integrate IPv6 in sensor nodes. In this concept, each node is addressed using IPv6, making it a true Internet of Things (IoT). Sensor nodes may be identified using the 6LoWPAN protocol established by the IETF [14]. The sink node encapsulates the detected data in IP packets before sending them to the Internet through the sensor overlay over IP as shown in Fig. 2a. The sink node is online, and the sensor nodes are virtualized in this design Fig. 2b. WSNs and the Internet are treated as different networks by using a higher-level gateway overlay. Both networks’ traffic is routed via the gateway to be redirected to the appropriate network. Internet routing information is transformed into WSN routing techniques by the gateway as in Fig. 2c. This is the most common way to link WSNs to the Internet. IP packet translation is the responsibility of the adapter at the sink node.

Convergence of Communication Technologies with Internet of Things

665

Fig. 2 Three schemes of connecting WSN with the IoT a the IP overlay over WSN, b the sensor overlay over IP, c the higher-level gateway overlays

6 Connecting MANET’s to IoT Mobile Ad hoc Network (MANET) is a self-configurable network of mobile nodes without the need for an infrastructure. These systems are often used in areas where infrastructure is scarce and patching it up is not an option. MANET has already shown its worth in disaster zones and on the battlefields. MANET’s multi-hop communication and low implementation costs make it beneficial in the above-mentioned locations. Every node in a MANET functions as a router, forwarding packets to the

666

V. Dankan Gowda et al.

next node on the network. Green communication, M2M, D2D, and the Internet of Things (IoT) are examples of recent research developments that demonstrate the value of ad hoc networks in reducing deployment and communication costs [11]. In order to link the MANET to the Internet, many technologies have been proposed. IP addresses are often assigned to nodes in a MANET so that data packets may be routed between them. As a result, it is always feasible to connect the MANET to the Internet. However, there are two major obstacles to overcome: (i) It is important for a node in a MANET to be able to quickly determine whether an address in the network is existent or not. (ii) Whether or not an access point or a gateway is required to connect to the Internet should be made very clear. Any node in the network may join or depart at any moment because to the nodes’ mobility and flexibility. This makes it difficult to gather neighbouring nodes’ IP addresses since nodes do not know where they are at any one moment. They use more time and message packets even if they are accessible. Connecting MANET to the Internet may be accomplished in a number of ways: Connecting a MANET to the Internet through an Access Point is the first option. However, the location of the access point is an issue. It is difficult to find the most efficient location for an access point or gateway. A mobile node might serve as a node of entry. Using two IP addresses, one for the MANET network and one for the Internet, is an alternative method for identifying nodes in the network. The target gateway, on the other hand, may be movable due to the nodes’ mobility. It’s possible that outbound connections might be disrupted if a node moves to a different gateway. Moving nodes may also be given IP addresses using the Dynamic Host Configuring Protocol (DHCP). Mobile nodes may now configure their IP addresses using this manner, however the DHCP server placement issue remains unresolved. As a result, MANET nodes may be automatically setup. In the literature, there is a lot of discussion on auto-configuration strategies. Connecting mobile nodes to the Internet via an appropriate auto-configuration approach in a MANET context is essential. From an IoT perspective, the capacity of nodes in a WSN to sense their surroundings and organise themselves into ad hoc networks to convey data is significant. However, three obstacles must be solved before the IoT may be used in more diverse ways [6]. Internet of Things (IoT) connectivity support for heterogeneous devices. In order for the Internet of Things to be a success, WSN nodes will need to have a number of common traits and functionalities. Sensor nodes’ battery power. IoT battery depletion and the necessity for a regular battery replacement pose a significant hurdle to wider deployments, despite many efforts to improve energy efficiency at different levels of the system. Nodes integrated with sensors and microcontrollers are too large to be used in an IoT system that will be available to everyone. To fully exploit the IoT’s potential, further progress must be made in the field of miniaturisation.

7 MANET-IoT Integration Protocol IoT’s WSN backbone requires two services from MANET: discovery and announcing. Discovery enables MANET nodes to look around the WSN topology

Convergence of Communication Technologies with Internet of Things

667

Fig. 3 MANET cluster formation

and choose a suitable node to connect to. WSN nodes are notified of the existence of MANET access points via the process of announcing. To prevent excessive power consumption, it is required to reduce the number of packet exchanges between MANET and WSN. Maintaining active communication and coordination with WSN packets for higher priority packets allows the MANET nodes to remain idle in typical scenarios. When a higher-priority data packet is detected in the WSN, any MANET node, which will be the cluster-head, initiates the cluster-formation procedure. As seen in the diagram, the procedure is divided into three sections. First, a MANET node identifies the higher priority WSN data packet as a cluster head and broadcasts a request to join the MANET group to other nodes. In the second phase, the MANET nodes that receive the cluster head’s request send the discovery message to the WSN nodes to locate an entry point for IoT and MANET connection. Nodes that receive and connect with IoT nodes might join the MANET cluster in the third phase. The NS-2 network simulator is used to test the proposed MANET-IoT integration protocol’s performance. MANET cluster formation is shown in Fig. 3.

8 Results and Discussion Because MANET nodes are distributed across a larger region and might be placed in places that are disconnected from the WSN, simulation findings demonstrate that increasing the number of MANET nodes has a significant impact on the delay of higher priority packets. Figures 4 and 5 shows this outcome. When MANET nodes move at 1.2 m/s, our protocol delivers over 90% of packets and drops over 60% of packets when nodes move at 12 m/s.

668

V. Dankan Gowda et al.

Fig. 4 Latency ratios between normal and higher priority packets

Fig. 5 The ratio of packets successfully delivered and the speed of MANET nodes

9 Conclusion Wireless sensor networks (WSNs), radio frequency identification (RFID), and mobile ad hoc networks (MANETs) are all discussed in this study (MANET). Though the Internet of Things (IoT) is interconnected via a variety of standards, protocols, and other means of communication, several issues remain. Further study is needed on energy management, energy efficiency, and energy-related assaults. IoT integration with MANET and WSN will help us build smarter environments because of the increasing prevalence of IoT services and applications in our daily lives. However, the IoT is also affected by the faults or challenges associated with these technologies. To overcome the problems and secure the Internet of Things (IoT) against different threats, new solutions must be developed.

Convergence of Communication Technologies with Internet of Things

669

References 1. Andrea I, Chrysostomou C, Hadjichristofi G (2015) Internet of things: security vulnerabilities and challenges. In: Computers and communication (ISCC), IEEE symposium. IEEE, pp 180– 187 2. Atakli IM, Hu H, Chen Y, Ku WS, Su Z (2008) Malicious node detection in wireless sensor networks using weighted trust evaluation. In: Proceedings of the 2008 Spring simulation multi conference. Society for Computer Simulation International, pp 836–843 3. Penna M, Jijesh JJ, Shivashankar (2017) Design and implementation of automatic medicine dispensing machine. In: RTEICT 2017—2nd IEEE international conference on recent trends in electronics, information and communication technology, proceedings, pp 1962–1966. https:// doi.org/10.1109/RTEICT.2017.8256941 4. Christin D, Reinhardt A, Mogre PS, Steinmetz R (2009) Wireless sensor networks and the internet of things: selected challenges. In: Proceedings of the 8th GI/ITG KuVS Fachespräch Drahtlose sensornetze, pp 31–34 5. Kishore DV, Shivashankar, Mehta S (2016) MANET topology for disaster management using wireless sensor network. In: International conference on communication and signal processing, ICCSP 2016, pp 0736–0740. https://doi.org/10.1109/ICCSP.2016.7754242 6. Varun CA, Shivashankar, Sahana M, Varun RS, Rajesh T (2018) Implementation of swarm intelligence in obstacle avoidance. In: RTEICT 2017—2nd IEEE international conference on recent trends in electronics, information and communication technology, proceedings, 2017, pp 525–528. https://doi.org/10.1109/RTEICT.2017.8256652 7. Sridhara SB, Naveen KB, Ramesha M, Pai GN (2020) Internet of things: internet revolution, impact, technology road map and features. Adv Math Sci J 9(7):4405–4414. https://doi.org/ 10.37418/amsj.9.7.11 8. Ramesh Naidu P, Guruprasad N (2020) Design and implementation of cryptcloud system for securing files in cloud. Adv Math Sci J 9(7):4485–4493. https://doi.org/10.37418/amsj.9.7.17 9. Isaiadis S, Getov V (2005) Integrating mobile devices into the grid: design considerations and evaluation. In: European conference on parallel processing LNCS, vol 3648. Springer, pp 1080–1088 10. Kasinathan P, Costamagna G, Khaleel H, Pastrone C, Spirito MA (2013) DEMO: an IDS framework for internet of things empowered by 6LoWPAN. In: Proceedings of the 2013 ACM SIGSAC conference on computer and communications security. ACM, pp 1337–1340 11. Ramesha M, Sridhara SB, Naveena Pai G (2020) FPGA implementation of low power high speed BTED algorithm for 8 bit error correction in cryptography system. Int J Emerg Trends Eng Res 8(7):3893–3897. https://doi.org/10.30534/ijeter/2020/158872020 12. Kolias C, Kolias V, Kambourakis G (2017) TermID: a distributed swarm intelligence-based approach for wireless intrusion detection. Int J Inf Secur 16(4):401–416 13. Lazarescu MT (2013) Design of a WSN platform for long-term environmental monitoring for IoT applications. IEEE J Emerg Sel Top Circ Syst 3(1):45–54 14. Mamun MSI, Kabir AFM (2012) Hierarchical design based intrusion detection system for wireless ad hoc network. Int J Netw Secur Appl (IJNSA) 2(3):102–117 15. Nadeem A, Howarth MP (2013) A survey of MANET intrusion detection and prevention approaches for network layer attacks. IEEE Commun Surv Tutorials 15(4):2027–2045 16. Reina DG, Toral SL, Barrero F, Bessis N, Asimakopoulou E (2013) The role of ad hoc networks in the internet of things: a case scenario for smart environments. In: Internet of things and inter-cooperative computational technologies for collective intelligence. Springer, Berlin, Heidelberg, pp 89–113 17. Schumacher C, Kushalnagar N, Montenegro G (2007) IPv6 over low-power wireless personal area networks (6LoWPANs): overview, assumptions, problem statement, and goals. Available: https://tools.ietf.org/html/rfc4919 18. Verma SS, Patel RB, Lenka SK (2015) Investigating variable time flood request impact over QOS in MANET. Procedia Comput Sci 57:1036–1041

670

V. Dankan Gowda et al.

19. Watteyne T, Vilajosana X, Kerkez B, Chraim F, Weekly K, Wang Q, Pister K (2012) OpenWSN: a standards-based low-power wireless development environment. Trans Emerg Telecommun Technol 23(5):480–493 20. Smys S (2019) Energy-aware security routing protocol for WSN in big-data applications. J ISMAC 1(01):38–55 21. Chen JI, Iong Z (2020) Optimal multipath conveyance with improved survivability for WSN’s in challenging location. J ISMAC 2(02):73–82

Chatbots: A Survey of the Technology Hrithika Singh, Asmita Bhangare, Rashmi Singh, Shubhangi Zope, and Pallavi Saindane

Abstract In recent years, chatbots have become more widely used in a variety of fields, including marketing, customer service, support systems, education, healthcare, cultural heritage, and entertainment. Live chat interfaces have gained popularity as a way to engage clients in real-time customer care in many e-commerce contexts. Artificial intelligence-based chatbots, commonly referred to as conversational software agents, are created to have natural language conversations with human users. They are progressively taking the place of human chat operators (AI). In this study, we’ve covered a chatbot’s fundamental definition as well as its numerous sorts and qualities. In this study, various methods for creating chatbots are also mentioned and described. Keywords Chatbot · AI—artificial intelligence · HCI—human–computer interaction · NLP—natural language processing · AIML—artificial intelligence markup language · RiveScript · ChatScript · NLU—natural language understanding · Rasa · ChatterBot · IBM Watson assistant · Dialogflow

1 Introduction Through the creation and examination of intelligent hardware and software, sometimes known as intelligent agents, artificial intelligence (AI) is gradually incorporating itself into our daily lives. The main motivation for doing research on this topic is that chatbots are trending in the modern world. The beginning of a new year is viewed as the start of new opportunities. Additionally, this suggests increased industry expansion, improvements to the current technology tools, and emerging AI technologies

H. Singh (B) · A. Bhangare · R. Singh · S. Zope · P. Saindane Vivekanand Education Society’s Institute of Technology, Chembur, India e-mail: [email protected] P. Saindane e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_49

671

672

H. Singh et al.

with greater potential. According to chatbot trends, more complex real-time interactions based on computational modeling will replace simple customer-focused inquiries. In the last few years, chatbots have grown rapidly and have succeeded in replacing various industries and internal positions in the company. So this paper focuses on the various types of chatbots, their trends, their characteristics, and all the other information one might need to deeply understand the use of chatbots. From simple labor to intricate operations, intelligent agents are capable of performing a wide range of tasks. One of the most fundamental and widely used types of intelligent Human–Computer Interaction is a chatbot, a type of AI system that is widely used (HCI) [1]. Computer programs called chatbots converse with users in everyday language. This technique, which dates back to the 1960s, sought to ascertain whether chat-bot platforms might deceive users into believing they were speaking to real people. A chatbot is defined as “A computer program designed to simulate interaction with human users, particularly over the Internet” [2]. Instead of offering direct touch with a real human agent, a Chatbot is a software application that conducts an online chat conversation via text or text-to-speech. In order to replicate a real-world conversation (or chat) with a user using messaging apps, websites, mobile apps, or the phone, chatbot systems use conversational artificial intelligence (AI) technology. It uses rule-based language applications to carry out live chat activities in response to user interactions occurring in real-time. Using Natural Language Processing (NLP) and sentiment analysis, it converses with users or other chatbots in human language via text or audio. In addition to chatbots, other names for them include interactive agents, intelligent bots, and digital assistants [2]. Virtual assistants and chatbots, which were formerly considered a niche concept in technology, have already become commonplace in India. Chatbots have primarily been relegated to consumer uses, according to cloud enterprise software vendor Ramco Systems. There are over 50,000 known bots on the market, the majority of which are geared toward consumers, and just a small percentage of them cater to business needs [3].

1.1 Chatbot Statistics In recent times terms like ‘chatbot’, ‘healthcare’, and their design have been a part of many Google searches by users. The table below shows the statistics of these searches through various databases. For example, the search string ‘Chatbot’ and ‘design’ has appeared 10,500 times by the users using the Google Scholar database as in Fig. 1. Figure 2 shows that after 2016, there was a noticeable growth in the use of chatbots. From 2000 to 2019, Scopus search results for the keywords “chatbot,” “conversation agent,” and “conversational interface” were organized by year [4]. The fastest-growing brand communication channel is chatbots: The use of chatbots as a brand communication channel has surged by a staggering 92% during 2019,

Chatbots: A Survey of the Technology

673

Fig. 1 Search results of chatbots

Fig. 2 Usage of chatbot as a brand communication

according to Drift’s 2020 State of Conversational Marketing study. In 2020, 24.9% of customers interact with firms via chatbots, an increase from 13% in 2019. Today’s consumers want 24 h customer service in many different industries, including banking and finance, and health and wellness. Due to this, companies are scrambling to create chatbots and virtual assistants that can respond to inquiries from clients at any time of day. Virtual agents are preferred by almost 40% of internet users, and as more and more important sectors like retail and healthcare adopt digital

674

H. Singh et al.

technology, chatbots are expected to gain more and more traction in the coming years [5]. A lot of businesses have seen significant chatbot growth, and 1.4 billion individuals currently utilize them on a regular basis [6]. Chatbots are increasingly prevalent on many websites and social media pages, despite the fact that the industry is still developing. 80% of internet users have interacted with chatbots at least once during that year, according to figures released by Userlike [7].

1.2 History of Chatbots Alan Turing suggested the Turing test in his 1950 work. If a panel of humans speaking with an unknown entity (via the keyboard, for example) thinks the entity is human yet the entity is a machine, the Turing test has been passed, according to the statement [2]. In 1966, the first chatbot ELIZA resembled the work of a psychiatrist, returning the user’s statements in the interrogative form [1]. PARRY was introduced in 1972, playing the role of a schizophrenic sufferer [1]. Rollo Carpenter developed the Jabberwacky chatbot in 1988 which intended to be entertaining while simulating a natural human discussion [8]. Dr. Sbaitso is a Chabot that Creative Labs developed in 1992 for MS-Dos. It was one of the first attempts to merge A.I. into a chatbot, and it is renowned for its fully voice-operated chat program [9]. A.L.I.C.E. is a chatbot that uses universal language processing and heuristic pattern matching to carry on discussions. It includes a web-based discussion feature that allowed for longitude and covered any subject but lacked any real awareness of the entire discourse [1]. The SmartChild, developed in 2001, was a predecessor to Siri in many aspects [9]. In 2010, Apple developed Siri for iOS. It has a user interface that uses natural language and is both an intelligent personal assistant and learning navigator. All further AI bots were made possible because of it [9]. IBM invented a chatbot called Watson in 2011. Watson was capable of comprehending natural human language [1]. Google Now was launched in 2012. It responds to inquiries, executes actions via requests to a set of web services, and gives recommendations. In 2014, Microsoft created the personal assistant Cortana. This program uses speech recognition and the proper algorithms to listen for and respond to spoken instructions [1, 9]. Google Assistant, which debuted in 2016, is the next version of Google Now. It boasts more advanced artificial intelligence and a nicer, more conversational interface, and it predicts user needs and delivers information to them [10–12]. Amazon’s Alexa is an intelligent personal assistant. It was launched in 2014 [9]. It is embedded into home automation and entertainment equipment, making the Internet of Things (IoT) more accessible to humans [13].

Chatbots: A Survey of the Technology

675

Early in 2016, a breakthrough in Artificial Intelligence Technology transformed the way people communicate with manufacturers. Developers were able to construct chatbots for their brand or service on social media platforms, allowing customers to complete certain everyday actions within their messaging apps [14].

2 Types of Chatbots The growth in chatbot technology is as dynamic as the evolution of chatbot capabilities. The types of chatbot is shown in Fig. 3. Chatbots are categorized based on their level of involvement and response generation mechanism.

2.1 Knowledge Domain Chatbots are categorized based on the amount of data they are trained on or the information they have access to Open Domain. These bots can converse about a variety of topics and answer accordingly. An open-domain dialogue system, as opposed to the traditional task-oriented bots, attempts to develop lasting relationships with users by fulfilling their need for social connection, affection, and communication. Meena, for example, is a Google open domain bot that can discuss anything Closed domain. These bots are focused on a specific knowledge topic, they may not be able to answer questions from other domains. A restaurant reservation bot, for example, will not inform you who America’s first black president was. For example, Domino’s ordering assistant bot as in Figs. 4 and 5.

Fig. 3 Types of chatbot

676

H. Singh et al.

Fig. 4 Meena

Fig. 5 Domino’s ordering assistant chatbot

2.2 Service Provided Based on the bot’s emotional proximity to the user they are classified into interpersonal, intrapersonal, and Inter-agent boats Interpersonal Bots. These include chatbots that operate in the area of communication that falls within the Proxemics chart’s Social or Personal distance range. These chatbots aren’t designed to be the user’s companions; instead, they’re supposed to collect data and send it on to the user; they’re just enablers. For example, FAQ bots Intrapersonal chatbots. These operate in the user’s domain and are typically seen in chat applications like Messenger, Slack, and WhatsApp and carry out tasks that are particular to the user’s domain. For

Chatbots: A Survey of the Technology

677

Fig. 6 Ask Sunny

Fig. 7 Cortana

example, Team Snapchat of Snapchat Inter-agent. This type of bot will be common in IoT-heavy locations. In this scenario, two systems communicate with one another to complete a task. For example, Inter-agent communication is demonstrated through the Alexa-Cortana interface as in Figs. 6, 7 and 8.

2.3 Goals Based on the primary goal, boats are classified into Informative boats, Conversational boats, and task-based boats Informative. These bots are intended to give the user data that has been saved or is available from a fixed source. For example, the IRCTC chatbot gives information about the services provided by IRCTC Conversational

678

H. Singh et al.

Fig. 8 Snapchat

Bots. These converse with the user as if they were another human being. Their purpose is to answer the statement that has been presented to them accurately. For example Vainu—Enrich customer conversations without form fill-ups Task-based. They specialize in one job, such as arranging a flight or assisting you with shopping. For example, H&M—Recommend your users based on their interests in Figs. 9, 10, and 11. Future of conversational AI. In the coming years, brands will continue to integrate AI and provide more meaningful customer service. Multilingual language capability and language switching are important priorities for brands if they operate from many regions or if the location has speakers of other languages. When it comes to customer service, emotional intelligence is a valuable asset. Conversational AIs are becoming more understanding of client intent, attitudes, and other emotions as a result of updated machine learning models. It enables them to respond more appropriately Fig. 9 IRCTC

Chatbots: A Survey of the Technology

679

Fig. 10 Vainu

Fig. 11 H&M

and gather crucial data. In the future years, AI may become self-aware enough to respond positively and perhaps create a welcoming environment. For years, these bots can remember past data, client emotional history during each engagement, and other sentiment-driven data. It can accurately select how to respond to the consumer and provide a safe atmosphere using this knowledge.

2.4 Method for Processing Input and Producing Responses Based on input processing and generation of replies, chatbots are classified into Intelligent Systems, Rule-Based Systems or Hybrid systems. To understand the

680

H. Singh et al.

Fig. 12 IBM Watson assistant

Fig. 13 ELIZA

inquiry, really Intelligent systems. They generate responses and use natural language comprehension. When the domain is narrow and there is plenty of data to train a system, these systems are used. For example IBM Watson Assistant Rule-based systems as in Fig. 12 and 13. These use Pattern matching, which makes them inflexible. When the number of possible outcomes is fixed and the number of situations is imaginable, these can be used. For example ELIZA Hybrid systems. These use a combination of rules and machine learning to achieve their goals. A system that manages the direction of conversation using a flow chart yet gives responses generated using natural language processing is an example [15].

3 Characteristics of Chatbot Chatbots have evolved to meet several business models, and they are currently used in a wide range of industries and sectors. The chatbot can perform a range of functions depending on its goal.

Chatbots: A Survey of the Technology

681

3.1 Conversational Development A great chatbot needs to have specialized natural language processing (NLP) skills to interpret and participate in conversations in several languages while also understanding the context of those conversations. To offer the appropriate first response, it may also identify the goal of a request and present alternatives to confirm or explain that purpose [16].

3.2 Autonomous Judgement The chatbot can carry out complicated thinking without requiring human intervention. For instance, an effective chatbot for customer support should be able to infer answers from pertinent case studies [16].

3.3 Omni Channel A chatbot is relevant enough to give users unified conversations across platforms or communication interfaces where the bot is connected. A chatbot can identify, recall, and relate to each client depending on their preferences, which are gleaned from past data, whether on the website or through a messaging channel [17].

3.4 Accessible All clients, regardless of language competence or any visual or hearing disabilities, must now be able to use chatbot interfaces equally. For example, bots should be able to read text aloud to the visually impaired [16].

3.5 Free to Explore Today’s chatbots may search the internet and company intranets for answers to client inquiries. As the customer knowledge base expands, the chatbot’s responses get better, and it might even figure out which articles were useful and which weren’t [17].

682

H. Singh et al.

3.6 Encourage Analytics A robust dashboard is offered by a time-relevant chatbot, allowing company leaders to monitor talks and progress by providing aggregate interaction data. Statistical information and performance evaluations are available to business owners, allowing them to enhance their bots and keep up their exceptional customer service [17].

3.7 Secure In the aftermath of recent data hacks, bots are safer than ever. Nowadays, a lot of businesses are quite transparent about the data they gather and how they use it. Users frequently have the choice to refuse data collection if they so choose [17].

3.8 Allows Upgrades Since developers are constantly trying to improve the operational abilities of various chatbots, a superb chatbot can upgrade without the need to remove the previous version before installing the new one. Unless an administrator user forces a modification, a standard chatbot upgrade does not result in data loss or changes in technical and functional [17].

3.9 Pre-trained The chatbot has already received training to identify brand- or industry-specific terminology. Even better, it’s already configured to handle typical customer demands in a certain sector [16].

3.10 Emotionally Intelligent The chatbot can understand sentiment and tone during a conversation to personalize the experience or escalate to a real agent if necessary. It can also deduce aspects of the client’s personality [16].

Chatbots: A Survey of the Technology

683

3.11 Integrates with CRM The chatbot may establish connections to crucial platforms and control workflows inside and outside the CRM. It can deal with everything, from a straightforward password update to a complex, multi-step workflow involving numerous programs, in real-time [16].

3.12 Sentences Are to the Point Sentences in good chatbot scripts are always short. Customers and consumers are prone to become overwhelmed when confronted with huge blocks of text and lose interest in the dialogue. In your script, removing needless words and using bullet points instead of long answers will assist maintain a brief and simple interaction and provide better customer service [18].

3.13 A Distinct Personality The script must consider what clients would find helpful, or motivating, just as it would in genuine interaction. The bot’s personality should be consistent with the brand’s. It’s just as crucial to stay in character as it is to get rid of the chatbot script’s bland and generic dialogue. This will make it easier to provide more effective real-time client service [18].

3.14 Handles the Unexpected The chatbot script is set up to deal with consumer misconceptions or ambiguities. Suggestions for similar inquiries or FAQs, as well as the preparation of an error message in the event of a misunderstanding, could be used to allow customers to return to the conversation without having to restart it [18].

3.15 Saves Money and Time Conversations that would otherwise necessitate an employee’s response can be automated using a chatbot, saving the company time and money. This time, money, and personnel can then be put to better use, resulting in improved outcomes [19].

684

H. Singh et al.

4 Chatbot Approaches, Tools and Architecture This section discusses the many methods of chatbot development, as well as the numerous tools available for chatbot development and, subsequently, the architecture of conversational bots.

4.1 Approaches Rule-based and machine learning approaches are two approaches to constructing a chatbot, depending on the algorithms and methodologies used. Rule-based Approach. The rule-based approach is presented and compared in the basic functionalities of AIML, RiveScript, and ChatScript. The pattern matching mechanism is commonly employed in chatbots. It creates effective and accurate responses for user input using pattern recognition or pattern matching. These languages are the most prevalent because they are lightweight and simple to configure. Machine learning-based approach. Machine learning-based approach because they examine the entire debate context rather than just the current turn, they do not require a predefined response for every possible user input. Natural Language Processing (NLP) is a branch of artificial intelligence that studies how computers use natural language text or speech. Keys used in NLP ChatBot: Utterance; Intent; Entity; Context; Session. The core of any NLP work is Natural Language Understanding (NLU). According to the user’s intent, NLU attempts to accurately extract context and meaning from unstructured natural language user inputs and responses. In other words, the intent is a relationship between what a user says and what action the chatbot should do.

4.2 Chatbot Tools Rasa is a Python-based framework for creating custom AI chatbots that understand natural language (NLU). ChatterBot is a Python package that makes creating automated responses to user input simple. Rasa is an AI chatbot development framework that employs natural language understanding (NLU). IBM Watson Assistant is a virtual assistant powered by IBM’s Watson AI Engine and available on IBM Cloud. Dialogflow is used by businesses such as Dominos, Shelf, Vonder, and SnapEngage.

Chatbots: A Survey of the Technology

685

Fig. 14 Architecture of chatbot

4.3 Chatbot Architecture Natural language understanding is a subfield of artificial intelligence that employs computer software to comprehend input in the form of sentences delivered via text or speech. When the Conversational Agent is in charge of a conversation, the DM system also acts as a state tracker, continuously maintaining the state of the conversation and triggering a transfer from one state to another. The NLU unit of dialogue systems generates the semantic representation for the dialogue task based on user input. The dialogue manager receives input from the NLU components, maintains some kind of state or interface with task managers, and then passes the output to the NLG. The architecture of chatbot is shown in Fig. 14.

5 Metrics for Evaluation of Chatbot This section explains how to test the performance of your conversational agent by using leading indications. An indicator that can be used to organize training to assist your chatbot to become a valuable team member.

5.1 Chatbot Activity This metric assists in determining the number of interactions, or the time it takes between a user asking a simple question and a constructive dialogue taking place [20]. This aids in determining the frequency with which chatbots are used and the number of users who use them [21].

686

H. Singh et al.

5.2 Chatbot Interaction Rate This metric aids in determining users’ attention during a conversation with a chatbot. It will enable you to determine the average amount of messages sent per chat [21].

5.3 Chatbot Session Length This metric aids in determining user interaction by comparing session length with other metrics. Session length provides additional information regarding conversational chatbot [22].

5.4 Chatbot Goal Completion Rate This metric aids in determining how many inquiries a chatbot must be asked before providing the required information. The higher the number of questions that users must ask, the longer it will take [22].

5.5 Chatbot Retention Rate This metric aids in determining how many people returned to utilizing the bot after a certain amount of time has passed. When compared to the normal frequency of client contact in your line of business, this is a significant difference. A high retention rate indicates that your strategy is working [22].

5.6 Chatbot Non-response Rate This metric aids in determining how many times the chatbot fails to answer a query. The chatbot often fails due to a lack of information [21].

5.7 Questions Per Conversation This metric aids in determining how many inquiries a chatbot must be asked before providing the required information. The more questions that users must ask, the longer it will take [21].

Chatbots: A Survey of the Technology

687

6 Comparison of Chatbots AI chatbots come in a variety of shapes and sizes. Some have clever artificial intelligence (AI), while others simply have branching questions with answers. Whether you’re utilizing an AI eCommerce chatbot or a simpler chatbot for assistance, the most important thing to consider is the user’s experience and whether they require an engaging discussion or a smart approach to fill out the information in a form or ticket. When a user requires additional assistance, live chat is frequently used to assist the dialogue [23]. The comparison on different chatbot is shown in Table 1.

7 Chatbots in Healthcare 7.1 Can We Trust Health in the Hands of Chatbots? Users’ readiness to disclose personal information, accept advice, and follow suggestions can be defined as trust in the context of health chatbots. The trust-building process is influenced by reliability, transparency, and explainability. The chatbots’ suggestions for recommendations must be consistent. The algorithms used by chatbots mainly rely on data. The data’s integrity, correctness, privacy, and security are critical. Users will not disclose sensitive healthcare information to use the health chatbots if this is not the case. Chatbots are enhancing productivity for healthcare staff and creating trust with patients in a future where virtual medical appointments are becoming the norm. Medical chatbots have the ability to benefit both patients and clinicians by lowering the burden and increasing the likelihood of a correct diagnosis. Trust generally starts where knowledge stops, and it acts as a means of bridging knowledge gaps [24].

7.2 Benefits of Chatbot in Healthcare By decreasing hospital visits, unnecessary treatments, and procedures, medical chatbots lessen the burden. Both time and money are saved. Doctor consultations are simple to schedule. By linking patients with the appropriate healthcare providers and assisting them in understanding their diseases and treatments without seeing a doctor, the chatbot reduces hospital readmissions. Additionally, medical chatbots are used by hospitals and private clinics to triage and clerk patients before they enter the consultation room. These bots use automatic responses to ask pertinent questions regarding the symptoms of the patient in an effort to provide the doctor with complete history [25].

688

H. Singh et al.

Table 1 Comparisions of different chatbot Title

Introduction

Pros

Ochatbot

Resolves inquiries and overcomes sales obstacles to engage consumers

Easy installation on At the time of posting, eCommerce platforms there is no SMS texting Strong reporting to or Whatsapp interface obtain insights and build business

Cons

ManyChat

Is powered by Facebook Instant Messenger, via which they communicate with consumers. As a Facebook Messenger widget, it has been added to websites

Visual dialogue builder for simple branching conversations Works on Facebook Does have SMS messaging

Limited to Facebook The conversation is restricted to a pre-defined branched conversations

Botsify

Botsify customer supports chatbots with Live Chat options to generate leads. Businesses can employ their customized bots for teaching, financial, and customer service purposes

Connect to live chat Connects to WhatsApp, Facebook Messenger, and SMS Messaging

The AI cannot simply transition from one interaction to another to gain marketing insights Cart reminders and notifications are not included

LandBot

Landbot is a chatbot platform that converts leads and collects data to generate leads. Landbot can be launched on platforms such as WhatsApp, Facebook Messenger, and websites

It has flow diagrams and many more customisable elements to match the demands of a business When the bot is unable to answer the customer’s questions, Landbot offers a function that allows the customer to connect with Live Chat

Landbot is not an AI chatbot and cannot change conversions Because it is mostly based on button clicks, it cannot engage in natural conversation like a flow bot No way for the user to type their question or search for products

Intercom

Is primarily a live chat system, but it is also evolving into a chatbot platform. Intercom’s Live Chats are well-known in the enterprise market, and they are now branching out into the chatbot industry

Switching languages within a discussion Features like targeted chats, reporting, and follow-up notifications are available Code-free personalization

They are gradually evolving into a more automated chatbot, however, this is not their primary focus when interacting with clients For live chat, the charge by the seat

(continued)

Chatbots: A Survey of the Technology

689

Table 1 (continued) Title

Introduction

Pros

Cons

Drift

Offers a pre-defined conversational experience that transitions to live chat for its consumers

One free seat on Live Chats’ customizable chat system is available. This includes unrestricted consumer contact

Live Chats occur in real-time. Unlike chatbots, having a human representative on the chat is required for 24/7 customer care

MobileMonkey MobileMonkey, which is Facebook Messenger, similar to Manychat and SMS, and WhatsApp are is powered by OmniChat all used to communicate technology, is frequently used for marketing, sales, and customer care

It does not work without Facebook, SMS, or WhatsApp. It is impossible to communicate with a customer who is not using these platforms

Tidio

Tidio offers clients a way to communicate with them via Live Chat and AI chatbots

Abandoned cart, order status, and product availability are all eCommerce features

The AI can’t hop from one chat to the next since it’s programmed to do so

Maisie

Maisie is a remarketing and sales-oriented chatbot platform

When a chatbot becomes Reporting insights are stuck or a live not present representative is requested, Maisie allows a live agent to take over the conversation

Chatfuel

Chatfuel uses Facebook Code-free building Messenger to automate conversations between their chatbot and humans

Chatfuel has no predetermined pricing because it is dependent on the number of users

8 Open Challenges Related to Chatbot The way that customers communicate with businesses has been changed by chatbots. As a result, the vast majority of businesses have jumped into the race to develop or enhance these virtual agents on their websites. But the developers must work around the obstacles that they present. The first issue is context in chatbots. The integration of context and meaningful responses is essential to the evolution of any chatbot because a conversation without context would be ambiguous. Limited User Attention is another problem, as users only have a short attention span and demand quick responses. Another major problem that contains the majority of the complexity is chatbot testing. Because natural language models are being improved, chatbots are always changing. Thus, testing and using chatbots becomes essential. All of your web data is often secure, but if you add specific chatbots to it, you can’t be sure whether or not the API will be secure. If appropriate security measures are not followed, data leaks and hacks are likely to occur. In addition to being equipped with the capacity to

690

H. Singh et al.

contact your target audience through brand identity and awareness, chatbots should be enhanced with pertinent data.

9 Proposed System During the pandemic, the demand for healthcare facilities increased. India is densely populated and does not adhere to the WHO standard of one doctor per thousand people. This kickstarted the idea of creating a Disease Prediction System using a Chatbot. We built a website and integrated a chatbot that predicts the disease the user is suffering from, as well as additional information and doctor recommendations for treatment. We chose to build a healthcare system in which patients can diagnose their illnesses by reporting symptoms to a chatbot, and they will also receive all necessary healthcare services for treatment. The primary goal of our project is to help people diagnose their illnesses early on by using our chatbot. So that they can take the necessary precautions. Along with disease prediction, our system aims to provide recommendations for doctors and medical laboratories in their area and also connect patients and doctors virtually through video conferencing.

10 Conclusion As little human interference with technology use as feasible is the aim of our technological environment. Chatbots are more efficient than humans and can reach a wider audience. At the same time, they have the potential to develop into helpful instruments for acquiring information. Running customer care departments is quite cost-effective for them. We think that the information provided in this study about the core ideas behind chatbots is helpful. Chatbots can be better understood by users and developers, who can then learn how to build and employ them for the purposes for which they were designed. Unquestionably helpful are chatbots. Healthcare professionals can use them on their websites to assist patients to get prompt medical information, booking appointments, sending daily reminders, and even issuing invoices. It might be impossible to distinguish between a chatbot and a real-life agent as AI and machine learning improve. Acknowledgements Nothing in this world is accomplished single-handedly. Every success is a result of much-needed guidance and support. We would like to thank all those people whose guidance and support have helped me make this paper. Firstly, we would like to thank our mentor Mrs. Pallavi Saindane who laid down the guidelines for this project, without which we would not have been able to complete this paper. Secondly, we would like to thank our parents and our siblings, who encouraged our work till the very end. Last, but not least, we would like to thank our dear friends, who have supported us in different parts of this project.

Chatbots: A Survey of the Technology

691

References 1. Adamopoulou E, Moussiades L (2020) An overview of chatbot technology. In: Maglogiannis I, Iliadis L, Pimenidis E (eds) Artificial intelligence applications and innovations. AIAI 2020. IFIP advances in information and communication technology, vol 584. Springer, Cham. https:// doi.org/10.1007/978-3-030-49186-4_31 2. Khanna A, Pandey B, Vashishta K, Kalia K, Bhale P, Das T (2015) A study of today’s A.I. through chatbots and rediscovery of machine intelligence. Int J u- and e-Serv Sci Technol 8:277–284. https://doi.org/10.14257/ijunesst.2015.8.7.28 3. Analyticsindiamag. https://analyticsindiamag.com/is-chatbot-a-boon-or-bane-heres-why-com panies-are-using-them-despite-tech-glitches/ 4. ScienceDirect. https://www.sciencedirect.com/science/article/pii/S2666920X21000175#fig2 5. Insiderintelligence. https://www.insiderintelligence.com/insights/chatbot-market-stats-trends/ 6. ChatBot. https://www.chatbot.com/blog/chatbot-statistics/ 7. Bloggingwizard. https://bloggingwizard.com/chatbot-statistics/ 8. Daffodil. https://insights.daffodilsw.com/blog/the-history-and-evolution-of-chatbots 9. Onlim. https://onlim.com/en/the-history-of-chatbots/ 10. ChatterBot. https://chatterbot.readthedocs.io/en/stable/ 11. Machine learning chatbot. https://neptune.ai/blog/building-machine-learning-chatbots-platfo rms-and-applications#:~:text=What%20is%20a%20machine%20learning,Natural%20Lang uage%20Processing%20(NLP) 12. Harms J-G, Kucherbaev P, Bozzon A, Houben G-J (2019) Approaches for dialog management in conversational agents. IEEE Internet Comput 23(2):13–22. https://doi.org/10.1109/MIC. 2018.2881519 13. Towards data science. https://towardsdatascience.com/5-reasons-why-your-chatbot-needs-nat ural-language-processing-ed20fb0a3655 14. GeeksforGeeks. https://www.geeksforgeeks.org/chatbots-using-python-and-rasa/ 15. Nimavat K, Champaneria T (2017) Chatbots: an overview types, architecture, tools, and future possibilities 16. IBM. https://www.ibm.com/blogs/services/2020/01/15/seven-characteristics-of-a-great-cha tbot/ 17. Visight. https://www.visight.io/blogs/62 18. HappyFox. https://blog.happyfox.com/must-have-characteristics-for-an-effective-chatbot-scr ipt/ 19. Ascentspark. https://www.ascentspark.com/blog/article/6-characteristics-ideal-chatbot 20. Kulkarni P, Mahabaleshwarkar A, Kulkarni M, Sirsikar N, Gadgil K (2019) Conversational AI: an overview of methodologies, applications and future scope. In: 5th international conference on computing, communication, control and automation (ICCUBEA), pp 1–7.https://doi.org/ 10.1109/ICCUBEA47591.2019.9129347 21. Measuring chatbot effectiveness. https://corporate.livingactor.com/en/measuring-chatbot-eff ectiveness-kpi/ 22. Metrics for you to evaluate the success of chatbot. https://www.mjvinnovation.com/blog/8-met rics-for-your-chatbot/ 23. Ometrics. https://ochatbot.com/best-ai-chatbots/ 24. Petrova V (2020) Can we trust our health in the hands of chatbots?: An exploratory study investigating the effect of anthropomorphic design of e-Health chatbots on patients UX 25. Benefits of chatbot in healthcare. https://topflightapps.com/ideas/chatbots-in-healthcare/#3

An Improved Machine Learning Algorithm for Crash Severity and Fatality Insight in VANET Network S. Bharathi and P. Durgadevi

Abstract A vehicular ad hoc network (VANET) is a wireless network that connects a group of moving or stationary vehicles together. VANETs were primarily used to provide safety and comfort to drivers in automotive environments until recently. Clustering is an important concept in vehicular ad hoc network (VANET) where several vehicles join to form a group based on common features. Clustering increases the complexity of data. The assessment of road accident strategies in Machine learning is presented in this paper. A road collision is the most unwanted and unexpected occurrence that may happen to a vehicle, due to the fact that they happen regularly. The goal of this study was to investigate the correlation between the concentration of collisions and injury. There are several factors that influence crashes, including weather, road conditions, driver distraction, and misread vehicle signals. Keywords Vehicular Adhoc network · VANET · Machine learning · Accident data · Collision

1 Introduction Much accident occurs in everyday activities and the number of victims increasing day by day. In the transport system, road safety is a crucial component. many people are harmed or died in accidents. The majority of these accidents are caused by human fault such as over speeding, violating the traffic rules, not wearing seat belt, drunk driving, and distracted driving [1]. Accidents can take place in variety of situations, and there is no guarantee that they might stop in the future because several people are controlled, such as traffic conditions, lighting conditions, wet roads, the number of people [2] on board, and the speed at the time of the collision. Figure 1: S. Bharathi (B) · P. Durgadevi Department of Computer Science and Engineering, SRM Institute of Science and Technology, Vadapalani Campus, Chennai 600026, India e-mail: [email protected] P. Durgadevi e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_50

693

694

S. Bharathi and P. Durgadevi

Fig. 1 Fatality statistic of Indian accident

Fatality Statistic of Indian accident In India, primarily cop officials gather, organize, and record accident data, therefore there is much scope for further improvement. They take extremely basic information and provide it in a non-analytical style for investigation reasons, and different researchers have done data analysis on that sort of dataset, but it becomes very tough to analyze particular circumstances.

2 Related Work Clustering is a major initiative for data analysis and an unsupervised learning method. Clustering entails putting together a collection of comparable items that are distinct from one another [1]. Clustering begins with defining the number of clusters, followed by determining the cluster’s centroid, and finally, ensuring that each cluster is the same size. In this study, a cluster of accident locations is created, which aids in the prediction and identification of risky places and scenarios in terms of road accidents. After that, take appropriate measures. Analyze as much data as possible and forecast important information [3–5]. The major goal of this study was to look at the elements that contribute to road Accidents and to forecast accident severity by applying Machine learning techniques. In this research, several classification algorithms were used to the accident visualization data set and the accuracy of three classifiers was compared with Decision tree, k-nearest neighbour, Naive Byes [6]. In this article, an analysis is carried out on incidents that occur at a given area on a regular basis or at the same spot every time [7, 8]. As a result, these types of data set analyses aid in the prediction of a given place, as well as the identification of characteristics that influence the likelihood of an accident in that region [9]. And for proceeding this, we used Kmeans algorithm for making a group of cluster about location and these clusters are divided into three different parameters depending upon the speed control such as low speed location, moderate speed cluster location and high speed cluster location [10–13] with this cluster classification the accident frequency count was then discovered as a metric to assist in the formation of location clusters. In this

An Improved Machine Learning Algorithm for Crash Severity …

695

paper, analyze about light conditions during accidents. we used K-means algorithm for making a light cluster and these clusters are alienated into five different parameters depending upon the lightening conditions of the road such as full day, Night with public lightening on, night without public lightening on, Twilight or dawn and night with public lightening not lit. This cluster of clusters provided a study of accidents that occurred more frequently during lightning conditions. This Study presents an overview of road accident analysis by using the methodologies of Machine learning to predict and prevent the accident. There are many algorithms available in machine learning for classification and clustering the data. Among those algorithms, Random forest were utilized in this study for accident analysis. The literature survey of the existing methods is shown in Table 1.

3 Crash Severity and Fatality Insight The severity of road accidents as defined by the number of people killed per 100 crashes went from 33.7 in 2019 to 36.0 in 2020. It emphasizes the importance of enhanced trauma healing and traffic-calming measures lowering the crash impact characteristics. Every year, over 1.3 lakh people are killed in road accidents in India, with over 3.4 lakh people wounded. The frequency of accidents and fatalities has decreased over time as a consequence of deliberate and coordinated road safety initiatives. We intended to explore the data to determine whether there are any factors that correspond to the number of injuries or fatalities recorded. Some data is accessible at https://morth.nic.in/state-wise-data and https://data.gov.in/. In this kernel, we will be studying road accident data from 2017 to 2019. Additionally discovered some statistical evidence for road accidents, such as weather conditions and road conditions. There is some more supplementary data available that you may utilize for better understand. The given Fig. 2 shows the analysis data set of Indian road accident data in state wise. India has the highest number of road fatalities in the world. In 2018 the latest year for which global figures are available, India accounted for more than a third of global road accident deaths. The World Health Organization says such deaths are under-reported and estimated that in 2016, the figure for India was likely twice as big as that reported by the government.

3.1 Accident Location Analysis The ultimate focus is to get latitude and longitude intelligence in order to contribute an immediate help that lowers fatalities. Clustering is a relevant data analysis venture and an unsupervised machine learning tool. Clustering entails putting together a

696

S. Bharathi and P. Durgadevi

Table 1 Overview of literature survey Proposed research

Technique used

Task/description

Pros

Cons

[1]

Decision tree, KNN

Univariate feature selection, and feature importance

Reduces overfitting

Ignore the interaction between classifier

[14]

XGboost

Speed limit, collision type, vehicle model and vehicle movement

It is very fast and accurate than Ada boost

It is sensitive

[15]

Association rule mining

Finds the rules to associate between object and set of item

Useful for analyzing and predicting user on VANET

Too many parameters used leads to high computation

[9]

k-mean and association rule mining

Make an array of regions to determine the co-relationship between accident factors

Useful in Computational analyzing data sets time is very high

[16]

Weka tool

Preprocessing, cataloging, grouping and visualization

Feature selection and data mining are integrated

It can handle only small datasets

[17]

IoT prototype, sensors

Detect accident and severity of the emergency level

Find out the location of the accident and nearest medical center. It sends a message for an ambulance after detecting basic information

All end user must be connected to the network

[18]

Self-organization map

Data are easily expounded and clarify

Grid clusters Neuron weights be observes necessary and similarities in data enough to cluster input

[19]

Navie bayes, Ada boost

Recursive feature elimination

Noise data and outliers are eliminated

Extremely sensitive

collection of comparable items that are distinct from one another [1]. Clustering begins with defining the number of clusters, followed by determining the cluster’s centroid, and finally, ensuring that each is the similar size. In this study, a cluster of accident locations is created, which aids in the forecast in addition identification of risky places and scenarios in terms of road accidents. After that, take appropriate measures. Analyze as much data as possible and forecast important information.

An Improved Machine Learning Algorithm for Crash Severity …

697

Fig. 2 Data set of Indian road accident data in state wise

The major goal of this study was to look at the elements that contribute to road. Accidents and to forecast accident severity by applying Machine learning techniques. In this research, several classification algorithms were used to the accident visualization. The accuracy of different classifiers and the data set were compared using Decision tree and k-nearest neighbor, Naive Byes [6]. In this article, an analysis is carried out on incidents that occur at a given area on a regular basis or at the same spot every time. As a result, these types of data set analyses aid in the prediction of a given place, as well as the identification of characteristics that influence the likelihood of an accident in that region [9]. And for proceeding this, we used K-means algorithm for making a collection of clusters about site and these groups are divided into three different parameters depending upon the speed control such as low, moderate speed, high-velocity cluster location. With these clusters classification the accident occurrence total was then discovered as a metric to assist in the formation of location clusters. Figure 3 given below discuss about the location analysis of accident. In accident location analysis, we taken different parameters such as Accident happened due to lighting conditions, weather conditions, lane analysis, over speed.

698

S. Bharathi and P. Durgadevi

Fig. 3 Location analysis

3.2 Accident Happened Due to Lighting Conditions This study analyzed about light conditions during accidents. We used K-means algorithm for making a light cluster and these clusters are alienated into five different parameters depending upon the lightening conditions of the road such as full day, Night with public lightening on, night without public lightening on, Twilight or dawn and night with public lightening not lit. This cluster of clusters provided a study of accidents that occurred more frequently during lightning conditions. Besides that, we used the characteristics in our dataset to try to figure out what went wrong the accident. In Fig. 4, statistically we have found some feature conditions where the number of accidents gets increased. Figure 5 illustrates several statistically significant situations that lead to an increase in the number of accidents.

Fig. 4 Road condition analysis

An Improved Machine Learning Algorithm for Crash Severity …

699

Fig. 5 Accident happened because of lighting

This was followed by a thorough examination of the lighting conditions in the areas where accidents were most common. 82.6% of accidents occur on a daily basis on average, and when public lights are on at night, the average accident rate is 9.42%.

3.3 Accident Lane Analysis We performed a test on where most of the accident happened and the type of lane based upon few characteristics using linear regression. The graph in Fig. 6 given below shows the analysis that on straight road the accident level is very high and the least accident count is presented on emergency stop band. In lane analysis, the accident rate is at very rate as 76.30% due to speeding and the drivers misleading transport constraints as shown in Fig. 6. Fig. 6 Lane analysis

700

S. Bharathi and P. Durgadevi

3.4 Analysis on Accident Happened Due to Weather The fact is that bad weather significantly increases the chance of accident. Conditions such as fog, light rain, cloudy weather, heavy rain, dazzling weather, fog smokes and other weather conditions as in Fig. 7. The chart given Fig. 8 shows the analysis of accident weather based on few features. The most accident happened on the normal day on an average 81.2%. Weather conditions contributing 81.2% in normal rate at high and contributes less in other conditions.

Fig. 7 Weather condition of fatality insight

Fig. 8 Accident weather analysis

An Improved Machine Learning Algorithm for Crash Severity …

701

Fig. 9 Accident caused by over speeding

3.5 Accident Caused by Over Speed Over speeding is a major factor in fatal accidents. To excel is part of the human mind as in Fig. 9. As long as man has a chance, he will attain infinite. But if we have to share the road with others, we’ll constantly be a few cars behind them. In the event of an accident, speed increases the chance of damage and severity of harm. In addition, quicker cars are more likely to be involved in an accident than slower vehicles, and the severity of the collision will be greater in the case of the faster vehicles [18]. While a slower vehicle will stop instantly, a quicker one will take a long time to stop and skid over a long distance because to the law of idea. Speeding vehicles are more likely to be involved in accidents and cause more injuries. Speeding also impairs the driver’s ability to predict what’s coming next, resulting in errors in judgment and, collision in the end.

4 Performance Analysis and Result In this research study to evaluate the performance of the offered techniques we conducted distinct experiments depending on the accident severity class. For accident severity classifications such as vehicle collision, speed, and lightning condition, the performance of each method has been determined. We observe that Decision tree and Random Forest algorithm works faster among other approaches. Table 2 shows the accuracy between the proposed approaches.

702 Table 2 Machine learning accuracy rate

S. Bharathi and P. Durgadevi Algorithm

Accuracy score

A-nearest neighbors

0.934712453515

Logistic regression

0.954389125349

Decision tree

0.968465327445

Random forest

0.9744156832658

5 Conclusion It is impossible to live with the consequences of traffic accidents in a growing country like us. To decrease the accident rate, in our nation, it has become necessary to regulate and organize road traffic using an unconventional structure. Road accidents can be avoided by taking simple precautions based on forecasting or notifications from a sophisticated system [16]. Moreover, it’s an urgent matter for our nation to report the fact that numerous individuals are killed in road accidents every day, and that the number is gradually growing. The use of machine learning is a reliable and effective approach to precise conclusion with the involvement to handle present circumstance, and the findings of the analysis section (Figs. 1, 2, 3, 4, 5 and 6) may be proposed to circulation powers that be for minimizing the sum of coincidences. We may utilise the recommended techniques to employ machine learning here since they have been proved to be more accurate in predicting traffic accident severity. By employing these techniques, we will also aim to create a recommender-system that can anticipate traffic accidents and alert road users. A mobile application based on this approach will be developed in the future to offer precise predictions for the user and make it extremely helpful.

References 1. Labib MF, Rifat AS, Hossain MM, Das AK, Nawrine F (2019) Road accident analysis and prediction of accident severity by using machine learning in Bangladesh. In: 7th international conference on smart computing and communications (ICSCC), pp 1–5. https://doi.org/10.1109/ ICSCC.2019.8843640 2. Mamun MAA, Puspo JA, Das AK (2017) An intelligent smartphone based approach using IoT for ensuring safe driving. In: International conference on electrical engineering and computer science (ICECOS), Palembang, pp 217–223 3. Rajkumar AR, Prabhakar S, Priyadharsini AM (2020) Prediction of road accident severity using machine learning algorithm. IJAST 29(06):116–120 4. Habibullah KM, Alam A, Saha S, Amin A, Das AK (2019) A driver-centric carpooling: optimal route-finding model using heuristic multi–objective search. In: 4th international conference on computer and communication systems (ICCCS), Singapore 5. Satu MS, Ahamed S, Hossain F, Akter T, Farid DM (2017)Mining traffic accident data of N5 national highway in Bangladesh employing decision trees. In: IEEE region 10 humanitarian technology conference (R10-HTC), Dhaka, pp 722–725 6. https://www.kaggle.com/phip2014/ml-to-predict-accident-severity-pa-mont

An Improved Machine Learning Algorithm for Crash Severity …

703

7. Verma V, Bhardwaj S, Singh H (2016) A hybrid K-mean clustering algorithm for prediction analysis. Indian J Sci Technol. https://doi.org/10.17485/ijst/2016/v9i28/98392,pp.0974-6846 8. Ma J, Ding Y, Cheng JCP, Tan Y, Gan VJL, Zhang J (2019) Analyzing the leading causes of traffic fatalities using XGBoost and grid-based analysis: a city management perspective. IEEE Access 7:148059–148072 9. Kumar S, Toshniwal D (2016) A data mining approach to characterize road accident locations. J Mod Transport. https://doi.org/10.1007/s40534-016-0095-5,pp.62-72 10. Princess PJB, Silas S, Rajsingh EB (2020) Machine learning approach for identification of accident severity from accident images using hybrid features. In: International conference for emerging technology (INCET), pp 1–4. https://doi.org/10.1109/INCET49848.2020.9154079 11. Patil J, Prabhu M, Walavalkar D, Lobo VB (2020)Road accident analysis using machine learning. In: IEEE Pune section international conference (PuneCon), pp 108–112.https://doi. org/10.1109/PuneCon50868.2020.9362403 12. Kin T, Goto J, Oshima M (2019) Machine learning approach for gamma-ray spectra identification for radioactivity analysis. In: IEEE nuclear science symposium and medical imaging conference (NSS/MIC), pp 1–2. https://doi.org/10.1109/NSS/MIC42101.2019.9059618 13. Paul J, Jahan Z, Lateef KF, Islam MR, Bakchy SC (2020) Prediction of road accident and severity of Bangladesh applying machine learning techniques. In: IEEE 8th R10 humanitarian technology conference (R10-HTC), pp 1–6. https://doi.org/10.1109/R10-HTC49770.2020.935 6987 14. Ting CY, Tan NYZ, Hashim HH, Ho CC, Shabadin A (2020) Malaysian road accident severity: variables and predictive models. In: Alfred R, Lim Y, Haviluddin H, On C (eds) Computational science and technology. Lecture notes in electrical engineering, vol 603. Springer, Singapore, 2020 15. Donchenko D, Sadovnikova N, Parygin D (2020) Prediction of road accidents’ severity on Russian roads using machine learning techniques. In: Radionov A, Kravchenko O, Guzeev V, Rozhdestvenskiy Y (eds) Proceedings of the 5th international conference on industrial engineering (ICIE 2019). ICIE 2019. Lecture notes in mechanical engineering. Springer, Cham 16. Gothane S, Sarode MV (2016) Analyzing factors, construction of dataset, estimating importance of factor and generation of association rules for Indian road accident. In: IEEE international conference on advanced computing, pp 15–16. https://doi.org/10.1109/IACC 17. Khaliq KA, Qayyum A, Pannek J (2017) Prototype of automatic accident detection and management in vehicular environment using VANET and IoT. In: 11th international conference on software, knowledge, information management and applications (SKIMA) 18. Al Malki A, Rizk MM, Mousa AA (2016) Hybrid genetic algorithm with K-means for clustering problems. Open J Optim 5:71–78 19. Sahin ¸ DÖ, Sirin ¸ B, Akleylek S, Kılıç E (2018) Work accident analysis with machine learning techniques. In: 3rd international conference on computer science and engineering (UBMK), pp 215–219. https://doi.org/10.1109/UBMK.2018.8566564

Network Monitoring of Cyber Physical System Mayank Srivastava, Aman Maurya, Utkarsh Sharma, and Shikha Srivastava

Abstract In the field of network and server security, every organization, no matter how big or small, has to adapt different monitoring techniques in relation with the servers and networks. Server attacks are a major threat in today’s world if the systems and networks and the host servers are vulnerable. Network and server monitors or administrators have to keep an eye over different tools for different purposes. Therefore, this paper offers a methodology with the help of which we propose to develop a tool that would ease up the work of server administrators. This tool would provide all the services that multiple other tools provide. This would help enhance the monitoring section of the web servers in particular by using different tools and modules with the help of python. This paper also explains and concludes that the tool being developed would be effective and efficient, and therefore save a lot of time and ensures security from threats related to servers. Keywords Server monitoring · Network monitoring · CPU · Wireshark · Network mapper (Nmap) · Memory consumption · Scapy

M. Srivastava (B) · A. Maurya · U. Sharma Department of Computer Engineering and Applications, GLA University, Mathura, UP 281406, India e-mail: [email protected] A. Maurya e-mail: [email protected] U. Sharma e-mail: [email protected] S. Srivastava Department of Mathematics, Aligarh Muslim University, Aligarh, UP 202002, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_51

705

706

M. Srivastava et al.

1 Introduction Cyber Security Incidents—including Sophisticated Cyber Security Attacks—can and do occur in many different ways. When we talk about risk to an organization due to cyber-attack, chances are, the organization can lose a significant amount of data in no time, affecting the speed and performance and majorly, the reputation of that particular organization. Therefore, to deal with such suspected or actual cyber security attacks or threats to become incidents, we need to record the events related to it, monitor them in a refined way on continual basis so as to investigate a suspected cyber security breach thereby, reducing the chances of a loss of data, information, etc. It will also remediate the damage easily. In this paper, an approach is presented in such a manner that it would make use of network packet sniffing, tracking, analyzing and extracting information, an easy task to perform. We have clustered modules such as wireshark, nmap, etc. with the help of python by creating a program that would run these modules in the background, which can be monitored effectively and efficiently. Section 2 includes related works of other researchers and authors in this field with their valuable contribution. Section 3 deals with the theoretical aspects of the background study of this paper and related topic’s discussion. Section 4 explains different modules and commands used in this work and their demonstration. Section 5 is dedicated to methodology used in this paper and Sect. 6 is dedicated to represent the obtained results. Section 7 gives the importance of the proposed approach. The conclusion is given in Sect. 8.

2 Review of Literature The area of Cyber Physical System is one of the ongoing technological research domain in which good number of researchers are working. Some of the major contribution of this area are as follows. Zeng and Wang [1] introduces how to monitor numerous servers via a very unpretentious protocol called SNMP. They used multi-threading technological aspect for the collection and processing of data thereby improving the efficiency of the collection. Roblee and Berk [2] developed a system named PQS which enables user-space monitoring of servers and makes accurate and fast decisions regarding server and service state. Yucheng and Yubin [3] focuses mainly on real-time monitoring and as it is very difficult to understand the software performance related issues by historical data analysis and remote monitoring. Forrest et al. [4, 5] showcases an initial result which focus at creating a different definition for UNIX processes which elaborates synonymous treatment along with normal behavior for itself. Bohra et al. [6] proposed a remote monitoring technique and recovery of the state of the software with taking no help from the processors and OS resources of the of a computer system. Kephart [7] proposed an autonomic computing that has a grand-challenge revelation of the

Network Monitoring of Cyber Physical System

707

future in which calculating systems will accomplish themselves according to the high-level purposes quantified by humans. Tsoa et al. [8] proposes an extensive survey on the management of server and network resources over Cloud infrastructures by highlighting key concepts and critically discussing their limitations and implications. Thirukonda and Becker [9] gives an automated architecture named WebSpy used to notify and take appropriate action when server downtime is identified. Sihyung et al. [10] shows the study of present network-monitoring technologies and identify different problems and suggested future directions. Suri and Batra [11] focusses on comparative study of different packet analyzers and criteria for choosing among them. Trimintzios et al. [12] presents the design and implementation of an API named DiMAPI used for managing flow creation and manipulation over distributed passive network monitoring. Fang et al. [13] proposed a method based on IA to complete dynamic network monitoring. The proposed method is developed using cross-platform language and includes automatic and manual mode. Bonelli and Giordano [14] presents Linux based approach named as PFQ that primarily fine-grained distribution to network applications and physical devices. Bashar and Smys [15] introduces an analysis of eavesdropping security issue in WSN environment. The paper proposed the notion of decreasing probability of interception and secure connection between the nodes. Haoxiang and Smys [16] discusses a concept based on optimization of energy related to Cyber-physical systems along with memory aware scheduling strategy and algorithms.

3 Basic Notions This section presents the various basic notions related to the concepts discussed in this paper.

3.1 Networking Networking, often called “computer networking”, is a well-known term in the field of computer science and also in cyber security domain. Usually understood as a practice, it involves moving and switching data between numerous nodes over a public medium of a statistics system. Network comprises mainly design, construction, operation, maintenance and management of the network. The level of expertise vital to operate a network, associates directly to the intricacy of that particular network which is under supervision. Computer networking permits all the devices and their endpoints to each other on a LAN (Local Area Network) or WAN (Wide Area Network)-for larger network. Networking includes everything such as telephone calls, texts, streaming of videos, internet and also, internet of things (IoT).

708

3.1.1

M. Srivastava et al.

Types of Networking

• Wired Networking: It necessitates the usage of a physical mode of transportation of data between two or more nodes. For example, copper Ethernet cables, optical fiber, etc. • Wireless Networking: It makes the use of radio waves and frequencies to transport data by air, empowering devices to be linked to a network without any cables. For example, microwave transmission, satellite, cellular and Bluetooth, wireless LAN, etc. 3.1.2

Some Basic Concepts

• TCP SYN Request: It is a form of denial-of-service attack in which an attacker hastily recruits a connection to a server without concluding the connection. The server has to spend resources to come up for half-opened connections, which can consume enough resources to make the system impassive to authentic traffic. • TCP Reset Connections: It represents an unforeseen closing of the session. It causes the resources assigned to the connection to be instantly released and all other information about the connection is removed. TCP reset is symbolized by the RESET flag in the TCP header set to 1. • TCP Half-Open Connections: The term half-open refers to TCP connections whose state is out of harmonization between the two interactive hosts, possibly due to a clatter of one side. A connection in the process of getting launched is also known as embryonic connection. • HTTP GET Requests: HTTP GET request method is used to regain data from a stated URL. The GET is the most popular HTTP request technique. GET requests should only collect data and should not disturb the state of the server.

3.2 Networking System performance monitoring is a method of collecting and analyzing the performance parameters of a system such as memory usage, I/O, CPU usage summary of node, etc. It involves presenting the data in such a way that it can be easily and effectively administered and understood by an administrator. Such services are very important and critical for the stable working of large clusters as it gives a green light to the administrator to investigate and find possible problems well before damage occurs. Also, other system software parts are benefitted with this information provided by the administrator after monitoring.

Network Monitoring of Cyber Physical System

3.2.1

709

Server Monitoring

In our daily lives, we have witnessed that computer network and size of communication has increased in a rapid fashion. As with this kind of pace, network has become a very important and irreplaceable asset for all of us. Therefore, network monitoring is as essential, which is evident from the fact that every organization now focuses on monitoring its network. Majority of all the network management software are mainly focused on the particular link and network equipment only, and not on server. As it appears, servers should be taken into consideration much like the networks or more. Therefore, this paper gives a brief explanation on how to monitor not only the networks and links, but also the servers and information related to it by developing a systematic monitoring tool. Main objectives to monitor server performance are as follows: • • • •

Server availability can be monitored. Responsiveness of the server is easily monitored. Server capacity and speed are also monitored effectively. Detection and prevention of issues that might affect the server proactively.

3.2.2

Network Monitoring

Network monitoring involves discovering, mapping and monitoring a computer network in a quest to ensure optimal performance and availability. Therefore, the only way to learn about what’s on a network is by monitoring it with the help of various network monitoring tools. The reason why we monitor a network is because it is the lifeline of IT infrastructure and almost all the critical information floats over some network. If there is any problem arising on a network due to a possible attack or malfunctioning, that critical information comes under a huge risk. A network monitoring system usually performs the following basic functions: • • • • •

Discovering Mapping Monitoring Alert Reporting.

Network administrators ensures everything is working in its supposed manner so that there may be no malfunctioning or threat for a disrupt. Routers, switches, ports, etc. usually comes under the scrutiny in case of any disruption. Hence, Administrator processes a report based on what is wrong and a critical action is then taken.

710

M. Srivastava et al.

4 System Modules In the process of developing the desired web server monitoring tool, we have studied many tools and techniques which are being used to monitor different aspects of a server and a network. Such tools, or we can say modules, includes wireshark, nmap, netstat command, scapy, socket, psutil, shutil and xampp stack package. Let us understand the basic functionalities of these modules and how we have compiled them based on their functionalities. Wireshark: It is an open-source packet analyzer, used for many purposes in the fields of communication protocol, network troubleshooting, software development, etc. to track the network packets in a manner as per need. Most commonly, Wireshark is considered as a sniffer tool or a network protocol analyzer. Wireshark also monitors unicast circulation which is directly not transmitted to the network’s MAC address interface, but it does various network taps which is also known as port mirroring to extend the capture at any point of time or traffic. The use of Wireshark in the implemented tool is shown in Fig. 1. How Wireshark command has been used in this tool • Wireshark captures packets, analyses them and displays every information related to it. Fig. 1 Use of Wireshark in the tool

Network Monitoring of Cyber Physical System

711

Fig. 2 Use of Nmap in the tool

• This information is stored in a.pkt file which is stored in the systems library with a certain number of packets and with different IP addresses as well. • We make use of these files to extract information about the running traffic on a network. Nmap: It is also an open-source utility for discovering a network. A Network Mapper (Nmap) is used to perform security auditing and scanning of a network. Its main use is to manage network inventory, monitor hosts and server uptime and downtime. We can use it over windows or linux or even Mac. Saving and comparing scan results is one of the main features of this tool. The use of Nmap in the tool is shown in Fig. 2. How nmap has been used in this tool • Nmap has a predefined task to map the network traffic by providing information regarding all the ports and services. • With the help of python and nmapscanner, we have collaborated nmap and our tool effectively to display all the information about the ports and services they offer. Netstat Command: Netstat Command is used in linux operating system which stands for “Network Statistics”. It provides information on different interface statistics, which includes open sockets, routing tables and connection information as well. Netstat command is also considered handy when used to display all the socket connections like TCP, UDP. It displays all those packets which are pending to get connections. How netstat command has been used in this tool • Netstat command is a replacement to many other statistics tools. • With just one command, we can capture live port analysis and information about the network we are operating at.

712

M. Srivastava et al.

Scapy: Scapy is a communicating packet handling program which is proficient for deciphering the packets of a wider quantity of protocols, while it sends them on the wire and captures them, makes ample number of requests and so on. The performance of scapy is commendable in its true sense for handling classical tasks like scanning, tracerouting, probing, unit tests, etc. Other specific tasks include invalid frames transfer. How scapy has been used in this tool • Scapy is a library in python which captures network packet. Therefore, the packets from wireshark that were first obtained are extracted using scapy into python. • We have run scapy using python on linux, but it can also run over windows, OSX and on most Unixes using libpcap. Socket: Sockets are basically the endpoints of a two-directional communication channel which communicates amongst a process or between two or more processes on the same system or between processes over different machines on different locations as well. We can implement sockets over different channel types, for example, Unix domain, TCP, UDP, etc. Its collection efficiently provides with precise classes for handling some public transports and generic interface to handle the rest others. Psutil: Also known as “Process and System Utilities”, Psutil is basically a crossplatform library used for retrieving information about running processes in python as well as system utilization like CPU, Memory, etc. The main use of Psutil is system monitoring, profiling and limiting process resources as well as management of running processes. How psutil has been used in the tool • Psutil utility modules has been integrated with python in the development of this tool in order to extract the information regarding the CPU usage. • CPU usage such as RAM usage, memory usage, etc. is extracted using pyautogui library. Shutil: Shutil is used to automate copying files and directories. Shutil helps in saving a lot of steps of opening, reading and writing and also closing of files, if there exist a condition when no actual processing is being done. This utility module is also used to achieve tasks like copying, moving or removing directory trees. How shutil has been used in the tool • Shutil has been used to extract information about the disk usage of the system being used for monitoring. • As we provide path of directory where python files are saved to shutil utility module, it extracts the disk usage and displays it. Xampp: Xampp is an open-source cross-platform web server solution stack package which makes transitioning from a local test server to a live server. It offers an apache application server, mariaDb database server, a php server, etc. and makes it easier to host. We used Xampp to host a dummy website’s application server and then ran all

Network Monitoring of Cyber Physical System

713

the monitoring tools with the help of a python concentric technique to get the desired outputs as part of our work. How Xampp has been used in our tool • Xampp offers many services like web server hosting, database server hosting, etc. We have used Xampp to host a web server that is to be monitored for various purposes.

5 Proposed Methodology Organizations bank on numerous servers such as core App servers, Database servers, Web servers, and caching servers for regular communiqué and serious business maneuvers. Servers are, undoubtedly, one of the most critical elements of the IT infrastructure as they’re largely used to bring about resources in a network. Therefore, we created a tool using python which monitors the performance of server and helps organization to proactively detect issues at early stages. We basically merged a few packet sniffing and tracing tools such as Wireshark, Nmap, etc. which gives the desired results as per our needs. To help matters, we constructed a python code to trace all the TCP packets using Wireshark. This helped us in getting information about what all packets are live on the server. Using Nmap, we tracked down the ports that we open on the server, in order to get information about the incoming and outgoing packets through those open ports. Secondly, server resource utilization is also observed using psutil and shutil modules with the help of python. It gives information regarding the CPU usage and RAM usage and also gives information about the disk usage statistics. Also, server uptime and downtime information are also gathered in the same manner. The basic process of implemented tool in terms of Use-case is shown in Fig. 3. In the build-up to this tool’s GUI, following are the key features of this tool: Features • It has four different buttons with their specified functionalities. • System-Server Information section provides all the information regarding the system and server. • Port status section gives a comprehensive information regarding all the ports open and closed on the server. • Packet analysis section delivers a detailed information regarding the packets transmitted, types of packets, flags used, etc. It also gives a count of total packets transmitted and total data transmitted in bytes. • Live Packet Analysis section captures live packet transmission. • Server Bandwidth section displays downloading and uploading speed on the server.

714

M. Srivastava et al.

Fig. 3 Use case diagram

6 Simulated Results The GUI of the tool—Web Server Monitoring is given in Fig. 4. Using the tool, we can monitor the server location, information of system and its architecture, bandwidth, uptime, CPU and RAM usage with disk usage and the server port status. The IP address of the server used—139.167.198.38. Monitoring Status “Connection Acquired” Connection acquired at: 2022-03-28 18:44:09 Monitoring started at: 2022-03-28 18:44:09 Disconnected at: 2022-03-28 18:44:017. System-Server Information Statistics Server Information: Location: 28.6519,77.2315, Region: Delhi, City: Delhi, Country: IN. Server System Information: System: Windows, Node Name: Utkarsh-Sharma, Release: 10, Version: 10.0.19044, Machine: AMD64.

Network Monitoring of Cyber Physical System

715

Fig. 4 GUI of the tool

Disk Usage Statistics: Disk usage statistics: usage (total = 250,025,603,072, used = 134,773,084,160, free = 115,252,518,912). The statistics is also shown in Fig. 5. CPU-RAM Usage: The CPU usage is 12.4% and RAM memory % used 55.6% The snapshot of the above statistics is also shown in Fig. 6. Server Port Status The port status of the Server is shown in Table 1. The snapshot of the tool which gives Server port status is given in Fig. 7. Fig. 5 Disk usage statistics

716

M. Srivastava et al.

Fig. 6 Server system ınformation

Table 1 Port status of server

Port

Status

Service

Port 21

Open

FTP

Port 22

Closed

NSH

Port 80

Open

HTTP

Port 139

Closed

Netbios-SSN

Port 443

Open

HTTPS

Port 8080

Closed

HTTP-alt

Packet Analysis The Server packet analysis is given below. The snapshot of this statistics is also given in Fig. 8. Total Packets Transmitted (In Bytes) 900 Total Data Transmitted (In Bytes) 226,850 IP Address—8.2.111.71 Total Packets Transmitted—19 Total Data Transmitted (In Bytes)—770 Percentage of Data transmitted to total 0.34 Total Data Sent (In Percentage)—53.25 Total Data Received (In Percentage)—46.75 Number of packets (Flag-wise #0 denotes Sent) 0A–10

Network Monitoring of Cyber Physical System

Fig. 7 Server port status

Fig. 8 Server packet analysis

A–9.

717

718

M. Srivastava et al.

Fig. 9 Server bandwidth status

Server Bandwidth The Server bandwidth status is given below. These statistics are also given in terms of snapshot in Fig. 9. Wifi Download Speed is 21734586.81508718 bytes per Second Wifi Upload Speed is 3065543.6396214203 bytes Per Second Therefore, this shows how the tool works where we don’t have to look around for multiple tools but a single platform is sufficient enough.

7 Importance of the Monitoring Tool Web Server Monitoring techniques represents the use of techniques and tools for security reasons. If you want to check what ports are open or closed, what services these ports are providing (in case you want to assess your server), then you have to use nmap for network traffic mapping. Similarly, if you want to go check how many data packets and of what type are travelling to and fro on your server, then you look for wireshark and so on. Therefore, this constant use of different tools and techniques takes a lot of time and complications. This paper comes up with a tool which is developed to give an efficient solution in this regard. Key important features of this project and monitoring tool are: 1. There will be no need of playing around with different tools and techniques.

Network Monitoring of Cyber Physical System

2. 3. 4. 5.

719

Web server monitoring becomes easier and efficient. With efficiency, time taken to check and analyze a web server is reduced. User friendly interface that provides easy and quick results. One domain and multiple services.

However, there are certain parameters that are still need to be focussed more precisely like data loss, throughput, efficiency etc. These parameters will be added as a part of future work of this paper.

8 Conclusion This paper gives a concrete solution as how we can manage different tools, techniques and modules in order to combine them significantly in a manner that every possible information about the server can be extracted easily and efficiently. With the help of python, we have achieved the objective of getting maximum information about the hosted server. Therefore, our objective is somewhat to develop a tool which can provide us with different information regarding the web server which helps ease up the monitoring when it comes to web server security. A significant action can be taken as all the information regarding the server is available on a single platform, rather than exploring different tools for different information. Further, the scope of these types of tools is huge. Time is money in today’s time and with increasing digitalization, we need to be very specific as to what need to be done to reduce the downtime. In this regard, this paper proposes efficient tool for server and network monitoring for reducing downtime.

References 1. Zeng W, Wang Y (2009) Design and implementation of server monitoring system based on SNMP. In: IEEE international joint conference on artificial intelligence, vol 3, pp 857–860 2. Roblee C, Berk V (2005) Implementing large-scale autonomic server monitoring using process query systems. In: IEEE proceedings of the second international conference on autonomic computing, pp 123–133 3. Yucheng L, Yubin L (2010) A monitoring system design program based on B/S mode. In: IEEE international conference on intelligent computation technology and automation, pp 184–187 4. Forrest S, Hoffmeyr S, Somayaji A, Longstaff T (1996) A sense of self for unix processes. In: IEEE symposium on security and privacy, pp 120–128 5. Nemeth E, Snyder G, Seebass S, Hein TR (1995) UNIX system administration handbook. Prentice Hall 6. Bohra A, Neantiu I, Gallard P, Sultan E, Iftode L (2004) Remote repair of operating system state using Backdoor. In: IEEE proceeding of the international conference on autonomic computing, pp 256–263 7. Kephart JO (2005) Research challenges of autonomic computing. In: Proceedings of the 27th international conference on Software engineering, IBM Thomas J. Watson Research Center, USA

720

M. Srivastava et al.

8. Tsoa FP, Jouet S, Pezaros DP (2016) Network and server resource management strategies for data centre infrastructures: a survey. Comput Netw 106:209–225 9. Thirukonda MM, Becker SA (2002) WebSpy: an architecture for monitoring web server availability in a multi-platform environment. Technical report CS-2002-07, Computer Science Department, Florida Institute of Technology 10. Sihyung L, Kyriaki L, Hyong SK (2014) Network monitoring: present and future. Comput Netw 65:84–98 11. Suri S, Batra V (2010) Comparative study of network monitoring tools. Int J Innovative Technol Exploring Eng (IJITEE) 1(3):63–65 12. Trimintzios P, Polychronakis M, Papadogiannakis A, Foukarakis M, Markatos EP, Oslebo A (2006) DiMAPI: an application programming interface for distributed network monitoring. In: Conference on network operations and management symposium, pp 382–439 13. Fang W, Zhijin Z, Xueyi Y (2008) A new dynamic network monitoring based on IA. In: IEEE international symposium on computer science and computational technology. IEEE, pp 637–640 14. Bonelli N, Giordano S (2016) Network traffic processing with PFQ. IEEE J Sel Areas Commun 34(6):1819–1833 15. Bashar A, Smys S (2021) Physical layer protection against sensor eavesdropper channels in wireless sensor networks. IRO J Sustain Wirel Syst 3(2):59–67 16. Haoxiang W, Smys S (2020) Secure and optimized cloud-based cyber-physical systems with memory-aware scheduling scheme. J Trends Comput Sci Smart Technol (TCSST) 2(3):141– 147

Impact of Security Attacks on Congestion in Wireless Sensor Networks Divya Pandey and Vandana Kushwaha

Abstract Security is an essential factor that must be addressed in wireless sensor networks (WSNs) since they are resource-constraint, may handle sensitive data, and operate in hostile inaccessible regions. The Purpose of security attacks in networks is mainly to breach the confidentiality, authenticity and, integrity of data which consequently raises the energy consumption rate and lessens the network lifetime. Several methods and techniques have been applied by attackers to achieve their goals. However, one of the easiest methods is to prompt congestion which causes packet delay, packet loss, re-transmissions and ultimately jeopardizes the network. Unfortunately, aggravation of network congestion by the faulty behaviour of malicious nodes has not been discussed to great extent in literature. In this paper, we explore and discuss different types of attacks to analyse their impact on congestion occurrence and related problems through simulation in different network scenarios. According to this analysis, different types of attacks could have devastating effects on networks. This demonstrates the need for developers to create more secure WSNs. Keywords Congestion · Security-attacks · Wireless sensor networks · Malicious nodes · Network lifetime

1 Introduction Researchers are paying a lot of attention to wireless sensor networks (WSNs) [1] due to their vast applications. Network security is one of the necessary requirements of these applications. However, WSNs are susceptible to various types of security attacks due to its stringent constraints and features [2]. Based on the target

D. Pandey (B) · V. Kushwaha Department of Computer Science, Banaras Hindu University, Varanasi, India e-mail: [email protected] V. Kushwaha e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_52

721

722

D. Pandey and V. Kushwaha

of attackers, security attacks in the network have been divided into two broad categories—Active and Passive attacks. An active attack disrupts the WSN by injecting defective data into it, impersonating, modifying resource and data streams, breaking security protocols, destroying sensor nodes, introducing bugs into the protocol, and overloading it with traffic. Rather than intercepting data communications, passive attacks focus on privacy (eavesdropping on data communications or monitoring packets within a WSN). Both types of attacks have adversarial effects on networks. As a result, security is a critical concern in WSN in order to maintain the confidentiality, availability, and integrity of sensed and sent data. Attacks on networks are primarily intended to deny the quality of service. Denying services in WSNs can be accomplished in several ways, including destroying packets and signals, congesting the network, dropping packets, or depleting the energy of network resources. Many congestion control algorithms for Wireless Sensor Networks ignore the impact of malicious nodes in network congestion as they presume that all nodes are authentic and operate properly, which is not a feasible assumption [3]. There can be various internal and external attacks in the network which might interrupt the usual operation of the network by compromising the transmitted data packets. Sensor nodes are prone to failure and malicious nodes aggravate congestion by sending multiple fake messages or dropping packets unnecessarily. Moreover, they can even generate heaps of packets abruptly. A network’s security can be improved by identifying the most harmful attacks that it can face. Keeping this fact in mind, in this paper, we explore and discuss different types of attacks that can cause congestion. Classification of these attacks has been done based on how they create congestion as shown in Fig. 1. Effects of security attacks on important network parameters such as PDR, throughput and network lifetime have been shown through simulation. In this way, theoretical as well as experimental aspects of security attacks are presented in this paper.

2 Related Works Security is one of the prime concerns due to its catastrophic consequences. As a result, it is one of the major research topics in the field of wireless sensor networks. There have been several attempts by researchers to explain what are the different kinds of attacks, as well as their detection techniques and countermeasures. A review of some of the relevant existing research articles in which security attacks in WSNs have been discussed in detail can be found below. Dewal et al. [4] have presented a series of challenges, security issues, and breaches in wireless sensor networks along with defensive measures that can be taken in response to these threats. Abasikele¸s-Turgut et al. [5] have investigated the effects of blackhole and sinkhole attacks on clustered WSNs using various performance metrics. Gavric and Simic [6] have discussed several common DOS attacks as well as potential prevention methods. They observed that DOS attacks can be divided into

Impact of Security Attacks on Congestion in Wireless Sensor Networks

723

A acks Causing Conges on

Packet Drop Method

Flooding Method

Black Hole A ack

DOS A ack

On-Off A ack

Hello and Session Flooding

Delay Method

Jellyfish A ack

False Route Method

Sinkhole A ack

Warmhole A ack

Jamming A ack

Fig. 1 Attacks causing congestion

multiple categories and they categorized DoS attacks based on the layers of protocol stacks. Diaz and Sanchez [7] have presented a methodology for security analysis in WSNs that involves attacker modelling and attack simulation, as well as performance analysis (estimation of node’s software execution time and power consumption). Following an examination of several WSN attack variants, an attacker model is developed. This model specifies three categories of attackers capable of simulating most WSN assaults. Tripathi et al. [8] have investigated the performance of the leach protocol in WSN when it was compromised by a black hole and a grey hole assault. They thoroughly explored different network characteristics with different node density. It has been noticed that the black hole attack has a greater impact on network performance than the grey hole attack. Suma [9] has analyzed the impact of the sybil attack on localization error in wsn and proposed a countermeasure by employing a detection and defence technique based on distance vector hop and the simulation result showed that proposed technique is able to decrease the average localization error buy a solid 4% when compared with previous methodologies. Vivekanadam [10] has presented a hybrid technique of hopfield neural network and firefly algorithm employing leach for preventing WSN from denial of sleep attack (DoSA) in which there is a significant loss of energy at the nodes due to their inability to enter sleep or power conservation modes. To address this issue, he

724

D. Pandey and V. Kushwaha

proposed a hybrid approach which results in a large improvement in network lifespan and energy usage patterns. Ganguly et al. [11] have proposed a novel Trust Integrated Congestion-aware Energy Efficient Routing algorithm (TCEER). They have calculated node potential using a Fuzzy Logic Controller based on its trust value, congestion condition, the remaining energy of the node, and distance from the current packet-transmitting node and the base station using a Fuzzy Logic Controller. In this way, they have combined security and congestion aspects together in their approach. The simulation results showed that security and congestion, when combined, yield positive network performance outcomes. Yan and Qi [12] a introduced congestion-aware routing algorithm (CARA). The method evaluates four route parameters: forward rate, node load factor, cache remaining rate, and forward average cache remaining rate, taking into account both geographical relationship and traffic load. Routing decisions are made using the multi-parameter fusion approach. As a consequence, the CARA algorithm recognises the sensor nodes and the surrounding area’s congestion perceptions and optimises network transmission performance. It is evident from the literature that although researchers have made significant efforts to provide security to WSNs against various kinds of attacks. Adversarial impact of these attacks on different network parameters have also been studied. However, congestion and security both aspects have not been given much consideration together, although they are somewhat inversely proportional. Therefore, in this paper efforts have been made to demonstrate the correlation between security attacks and congestion and it is observed that the impact of attacks on congestion is severe.

3 Attacks that Contribute to Congestion Occurrence Some internal attacks can cause congestion on sensor nodes, links or on both. There are variety of methods for attackers to make network congested. Therefore, depending on the approach of creating congestion these attacks have been classified into 4 major categories. A short description of these methods has been given in the following subsection.

3.1 Packet Drop Method The idea behind this type of attack is basically to destroy the packets or signal so, as to prevent them from reaching to the destination so that they must be retransmitted resulting in a huge number of packet drops and retransmissions. In this way, this method hampers QoS parameters such as packet delivery ratio, network throughput and end-to-end delay etc. Attacks that follow packet drop pattern are as follows

Impact of Security Attacks on Congestion in Wireless Sensor Networks

725

• Black hole Attack Blackhole attack tries to divert network traffic towards the malicious node by advertising zero or low cost towards the Sink node. Basically, it tricks neighbouring nodes into sending packets to the compromised node, where another attack takes place. The blackhole attack is usually combined with Selective Forward attack which discards some or all the packets it receives by neighbour nodes [9]. In worst case, a malicious node imitates a black hole, discarding all data packets that travel through it as matter and energy vanish from our cosmos. If the attacking node is a connecting node between two network connecting components, the network is essentially divided into two separate components. • On–Off Attack In On–off attack as its name suggests, malicious nodes might opportunistically act good or bad. Nodes in the network are compromised in such a way that bad behaviour would go unnoticed [10]. As a result, malevolent nodes can maintain their trust despite behaving negatively. So, this type of attack can only be identified based on the analysis of misbehaviour history. • Jamming Attack The purpose of jamming attack is to destroy a signal or a packet. The term “destroy” here refers to rearranging a signal or modifying a few bits of a packet by interfering with transmission [5]. Jammers are one of the earliest and most well-known WSN attacks. Jamming can occur at any OSI layer of a network.

3.2 Flooding Method In flooding technique attackers make lots of replica of packets and overwhelm the sensor node with huge number of incoming packets making it incapable of processing them further so it become congested. A list of flooding based attacks are as follows • Denial of Service (DoS) Attack A DOS attack seeks to interfere with the proper operation of a network.in DoS attack compromised node makes lots of replica of received packets and transmit it to its neighbour node. And the receiver node becomes unable to handle this multitudinous packet simultaneously so it gets congested [6]. Thus, DoS assault affects a target by flooding it with massive volumes of packets thereby degrading or rendering the target’s service useless. • Hello and Session Flooding The motive of this type of attack is to exhaust the node’s resources or drain its energy. Channels are continuously flooded with requests or transmissions from malicious nodes. Using a large transmission power, a network attacker could broadcast “HELLO” packets (used in many protocols for node discovery) in order to fool all nodes into thinking the adversary is within one-hop communication range, thus wasting excessive energy sending packets to an imaginary neighbour. Sensor nodes periodically send out a special data packet (message) called a HELLO packet to establish and confirm network relationships with

726

D. Pandey and V. Kushwaha

other sensor nodes. Once routers successfully exchange HELLO packets, they can automatically establish adjacency and can begin to route data between them.

3.3 Delay Method Motive of delay-based method is to cause high end-to end latency thereby reducing throughput and QoS. • Jelly fish Attack Jellyfish attack is similar to blackhole and grey hole attack but the difference is that in a jellyfish attack, the packet will first be delayed before and after transmission and reception in the network then it may drop packets also, whereas in a black hole attack the packet is only dropped. Jellyfish attack may also jumble the order in which packets are received before sending them in random order. The regular flow control method utilised by nodes for reliable transmission is disrupted as a result of this attack. A jellyfish assault can cause a long end-to-end latency, lowering QoS.

3.4 False Route Method False route technique, as its name implies, entails sending data packets over the routing path in the incorrect direction, rendering the destination unreachable. Adversary nodes can do this by disseminating fake routing information or overflowing routing table with routes that do not exist. In this section we have briefly explained these attacks which directly or indirectly give their contribution to congestion occurrence are listed below. • Sinkhole Attack During a sinkhole attack, an attacker compromises nodes in the network and advertises fictitious routing updates to lure network traffic towards the compromised node. It advertises fake messages saying it is the node with zero hop distance from the sink i.e., it imitates as sink in order to capture all the packets that were meant to receive by actual sink node [11]. A sinkhole attack can also be used to launch other attacks, such as selective forwarding, acknowledging spoofing, and dropping or altering routing information. A base station can also receive bogus information from it as the data packets can be modified by compromised node before relaying it to the actual sink node [12]. • Wormhole Attack The attack involves more than one malicious node and is carried out at the network layer. Compared to normal nodes, the nodes used in this attack are capable of establishing better communication channels over a wide area [13]. This attack involves tunnelling data between one malicious node and another at the other end of the network [14, 15]. Consequently, the other nodes in the WSN may believe

Impact of Security Attacks on Congestion in Wireless Sensor Networks

727

they are closer to other nodes than they actually are, causing routing algorithm problems. Compromised nodes may also interfere with data packets. It is also possible to combine wormhole attack with sinkhole attack in order to enhance its effectiveness [16, 17].

4 Experimental Results We have setup a WSN in 100 * 100 m2 region with 16 static sensor nodes that are randomly deployed and one sink node which is different from all the sensor nodes in terms of power, storage etc. We assume all the sensor nodes have equal initial energy, same hardware configuration and interface. For analyzing the impact of security attacks on congestion and associated network parameters such as packet delivery ratio (PDR), energy consumption, network lifetime etc. we have purposefully injected malicious nodes rendering different types of attacks into the network. The simulation also takes network deployment into consideration as attacks may have a different effect depending on the topology of the network, the software on the nodes, the hardware components or even the configuration of the nodes/networks. In this way, the WSN simulation will assist in identifying the most problematic attacks and the most vulnerablearts of the network. Table 1 shows the simulation parametrs and its corresponding values. Table 1 List of simulation parameters

Network parameters

Value

Network area

100 * 100 m2

Number of nodes

16

Initial energy of nodes

0.5 J

Round of simulations

200 rounds

Malicious nodes

1–4 nodes

Sink position

(50, 50), (50, 100)

Buffer size

10 packets

Data packet size

1024 Bits

Control packet size

200 Bits

728

D. Pandey and V. Kushwaha

4.1 Effect of Attacks in Different Network Scenarios In this section, we examine how attacks behave based on the distances from sinks and network topology. Experiment-1 In this experiment firstly, we simulated the network without attacking node, then we intentionally introduced malicious nodes launching attack into the network and gradually minimised its distance from the sink to see the effect of attack when malicious nodes approach the sink. During this test, it is observed that interference increases whenever compromised node approaches the sink, and that the number of packets that are received decreases significantly. Table 2 outlines the whole experiment scenario. Firstly, in the network without attack total 50 packets were sent and all were received that means no packet drops happend. When network is victimzed with attack, position of attacker node from sink node is gradually decreased and it was seen that as distance between sink node and attacker node reduces number of packet drop increases. Experiment-2 In this experiment we intend to analyse the behaviour of malicious nodes by changing the sink positions in the network or we can say by changing the network topology. We have taken two test cases; in first case we have deployed sink inside the sensing region and placed it in middle of the network area. (represented by Fig. 2a) and in second case sink has been placed farther from the sensing region (represented by Fig. 2b) thereby placed in center top position. And it is observed and shown in Fig. 3 that if sink is placed inside the sensing region it is less affected by some attacks as compared to sink which is placed farther from the sensing nodes, possible reason behind this could be that sensor nodes can find more alternate routes for packet transmission when sink is placed between the sensor nodes. The experiment involved transmitting 30 data packets to the sink when the network was free of attacks and when attacks were launched on it. And, it was witnessed that sink positions do not have any significant effect on the network in the absence of attacks as in both cases when (sink was postioned in middle and when sink was on top) almost all the data packets were successfully received by it. However, When the network contained malicious nodes, then a sink located inside the sensing region was less vulnerable to attacks than a sink located far away sensor nodes. Table 2 Impact of security attacks on network when it approaches sink Without attack

With attack Distance from sink 50 m

Distance from sink 30 m

Distance from sink 10 m

Total successful transmission

50

34

27

15

Total packet drops

0

16

23

35

Impact of Security Attacks on Congestion in Wireless Sensor Networks

729

Fig. 2 Sensor network topology

Impact of a acks on different Network Scenario 40 30 20 10 0 Packets Packet Received Packets Packet Received Forwarded to Sink by Sink Forwarded to Sink by Sink Without A ack When Sink in Middle

With A ack When Sink on Top

Fig. 3 Effect of attacks on different sink positions

4.2 Effect of Attacks on Important Network Parameters This section shows the impact of various types of security attacks on crucial network parameters such as packet delivery ratio, Throughput and Network lifetime with the help of graph. • Impact of Attacks on PDR “The packet delivery ratio (PDR) is defined as the ratio of total packets delivered to total packets sent (including RREP and RREQ) from a source node to a destination node in a network” [18]. The aim is to deliver maximum number of data packets sent to the destination. Figure 4 depicts what impact do the different

730

D. Pandey and V. Kushwaha

Impact of A acks on PDR 1 0.8 PDR

0.6 0.4 0.2 0 PDR A acks Without A ack

Black hole A ack

On-Off A ack

DoS

Sinkhole A ack

Warmhole A ack

Fig. 4 Effect of attacks on packet delivery ratio

Fig. 5 Energy consumption analysis

Energy Consump on in percent

types of attacks have on the value of PDR. Network with all the legitimate nodes get the average PDR value between 0.91 and 0.98 whereas in presence of attacks it gradually decreases. We have observed that black-hole attack shows the worst effect on PDR. • Impact of Attacks on Energy Consumption Some security attacks exploit WSN’s limited energy resource by replicating massive amounts of data packets or by dropping lots of data packets that require multiple retransmissions, so they increase network energy consumption per transmission, which results in a shorter network lifetime. In the following graph (Fig. 5) impact of attacks on energy consumption has been shown. As demonstrated by the experimental results, security attacks can have extensive adversarial effects on energy consumption.

Energy Consump on Analysis 100 50 0 1

21

41

61

81

101 121 141 161 181

Number of Rounds Without A ack

With A ack

Impact of Security Attacks on Congestion in Wireless Sensor Networks

731

5 Conclusion Some security threats have a direct influence on network congestion which brings more processing and communication overhead, as well as an increase in energy consumption, which essentially limits network lifetime. The majority of existing congestion management techniques for Wireless Sensor Networks ignore the impact of security threats by malicious nodes on network congestion. Malicious nodes exacerbate congestion by delivering fraudulent messages. This paper examines possible network attacks that are crucial to understand while developing congestion countermeasures. A systematic analysis of the different WSN attacks has been carried out and existing research results have been critically analysed in this study.

References 1. Zheng J, Jamalipour A (2009) Wireless sensor networks: a networking perspective. Wiley, Hoboken 2. Yick J, Mukharjee B, Ghosal D (2008) Wireless sensor networks survey. Comput Netw 3. Pandey D, Kushwaha V (2020) An exploratory study of congestion control techniques in wireless sensor networks. Comput Commun 157:257–283 4. Dewal P, Narula GS, Jain V, Baliyan A (2018) Security attacks in wireless sensor networks: a survey. In: Cyber security. Springer, Singapore, pp 47–58 5. Abasikele¸s-Turgut I, Aydin MN, Tohma K (2016) A realistic modelling of the sinkhole and the black hole attacks in cluster-based WSNs. Int J Electron Electr Eng 4(1):74–78 6. Gavric Z, Simic D (2018) Overview of DOS attacks on wireless sensor networks and experimental results for simulation of interference attacks. Ing Inv 38(1):130–138 7. Diaz A, Sanchez P (2016) Simulation of attacks for security in wireless sensor network. Sensors 16(11):1932 8. Tripathi M, Gaur MS, Laxmi V (2013) Comparing the impact of black hole and gray hole attack on LEACH in WSN. Procedia Comput Sci 19:1101–1107 9. Suma V (2021) Detection of localization error in a WSN under Sybil attack using advanced DV-hop methodology. IRO J Sustain Wirel Syst 3(2):87–96 10. Vivekanadam B (2020) A novel hybrid HNN and firefly algorithm to overcome denial of sleep attack on wireless sensor nodes. J Ubiquit Comput Commun Technol (UCCT) 2(04):223–227 11. Ganguly S, Chakraborty A, Naskar MK (2013) A trust-based framework for congestion-aware energy efficient routing in wireless multimedia sensor networks. arXiv:1312.4071 12. Yan J, Qi B (2021) CARA: a congestion-aware routing algorithm for wireless sensor networks. Algorithms 14(7):199 13. Gaware A, Dhonde SB (2016) A survey on security attacks in wireless sensor networks. In: 3rd international conference on computing for sustainable global development (INDIACom). IEEE, pp 536–539 14. Patel MM, Aggarwal A (2013) Security attacks in wireless sensor networks: a survey. In: International conference on intelligent systems and signal processing (ISSP). IEEE, pp 329–333 15. Kibirige GW, Sanga C (2015) A survey on detection of sinkhole attack in wireless sensor network. arXiv:1505.01941 16. Kavitha T, Sridharan D (2010) Security vulnerabilities in wireless sensor networks: a survey

732

D. Pandey and V. Kushwaha

17. Le A, Loo J, Luo Y, Lasebae A (2013) The impacts of internal threats towards routing protocol for low power and lossy network performance. In: IEEE symposium on computers and communications (ISCC). IEEE, pp 000789–000794 18. Pandey D, Kushwaha V (2019) Performance parameter analysis of congestion control in wireless sensor networks. In: 4th international conference on information systems and computer networks (ISCON). IEEE, pp 656–661

IoT Weather Forecasting Using Ridge Regression Model Karthik G. Dath, K. E. Krishnaprasad, T. S. Pushpa, and K. P. Shailaja

Abstract Weather Forecasting is an Internet of Things (IoT) based initiative that attempts to provide weather forecasting via the internet via a website. Using several sensors, our project Weather Forecasting collects data such as temperature, humidity, rain, and pressure. Our project also includes a facility for measuring atmospheric conditions to give information for weather forecasts and to study weather and climate. As a result, Weather Forecasting using the Internet of Things is proposed to assist consumers in accessing weather data anywhere in real-time. For the data storage, a real-time database has been used using Firebase and will be displayed in the dashboard which was designed using React JS. Keywords Internet of Things · Weather forecasting model · Regression model · Sensors

1 Introduction Weather forecasting is an application of science and technology to forecast weather conditions based on prior data. Weather forecasting has been attempted formally since the nineteenth century and informally for millennia. Weather forecasts area unit created by assembling quantitative knowledge on this state of the atmosphere, land, and ocean, and so applying meteorology to forecast however the atmosphere can modification at a particular place. K. G. Dath (B) · K. E. Krishnaprasad · T. S. Pushpa · K. P. Shailaja B.M.S College Engineering, Bangalore, India e-mail: [email protected] K. E. Krishnaprasad e-mail: [email protected] T. S. Pushpa e-mail: [email protected] K. P. Shailaja e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_53

733

734

K. G. Dath et al.

Our project Weather Forecasts is an Internet of Things-based project that intends to provide weather information and forecasting using the Ridge Regression (RR) Model, a machine learning method. Using sensors from the surrounding environment, our project captures data such as temperature, humidity, rain, and pressure. Our project also includes the ability to measure atmospheric conditions to provide data for weather forecasts. We used Firebase’s real-time database to store sensor data, and a ReactJS-based website to display information. We store the weather data in ThinkSpeak, which is an open-source IoT analytics platform. It visualises data and allows us to download the recorded data together with the timestamp. The API supports data retrieval and logging by connecting to devices and websites. It can also perform online analysis and processing of the data it comes from. Using the RR model, we use the acquired data to forecast the weather. As a result, we advocated that people have real-time access to weather information wherever they are. Also, because we collected data in our home setting, the forecasts may not be completely accurate.

2 Hardware Requirements Node MCU (ESP8266) as in Fig. 1 is a low-cost and powerful open-source microcontroller board which connects objects and lets data transfer using Wi-Fi protocol [5]. It is easy to use and can be programmed with Arduino IDE, it is available as an access point or station and also consist of an internal antenna. The absolute pressure of the air around Barometric Pressure sensors is measured as in Fig. 2. This pressure is affected by both the weather and altitude. Depending on how you interpret the data, you may monitor weather changes, measure altitude, or do any other operation that necessitates an exact pressure reading [6]. The ultralow power consumption down to 3 μA makes the BMP 180 the leader in power saving for the devices. BMP 180 is also distinguished by its very stable behaviour (performance) regarding the independence of the supply voltage [7]. The DHT11 in Fig. 3 is a low-cost sensor, used to measure temperature and humidity [9]. The DHT-11 Sensor is a straightforward, incredibly affordable digital Fig. 1 NodeMCU ESP8266

IoT Weather Forecasting Using Ridge Regression Model

735

Fig. 2 Barometric pressure (BMP) 180

thermometer. It measures the air around it using a capacitive humidity sensor and a thermistor, and it spits out a digital signal on the data pin (no analogue input pins are needed). The raindrop sensor in Fig. 4 detects rain. It has two modules one that detects rain and another the control module that compares and converts analogue values to digital values. The analogue output is employed as a section of discovery of drops within the life of rainfall related to three 3V/5V power provided and the device works in lightweight of the amount of the water interfacing the raining board, the output voltage of the gadget varies on the length of the raining board being wet that is modified over to digital through ADC chip [8]. Fig. 3 Digital humidity temperature (DHT 11)

736

K. G. Dath et al.

Fig. 4 Raindrop sensor

3 Software Requirements Arduino IDE: Arduino Software (IDE) is an open-source editing software for writing code and uploading code to any compatible Arduino microcontroller board. Arduino board can be also used with other third-party boards by installing appropriate libraries [1]. ReactJS: ReactJS is a JavaScript library for creating user interfaces, single-page applications, and reusable UI components [2]. Firebase: Firebase is a Google back-end and cross-platform application development that helps you build and deploy. Firebase provides services like real-time database, storage, hosting, authentication, function and ML. In our project, we have used Hosting and real-time database services. [3]. Jupyter Notebook: it is an open-source tool for writing and execution of the program. It is used to create and share documents that contain live code, equations, visualizations, and text [4]. ThinkSpeak: It is a cloud-based IoT platform for the analysis and visualization of IoT data. It visualises data and allows us to download the recorded data together with the timestamp. The API supports data retrieval and logging by connecting to devices and websites. It can also perform online analysis and processing of the data it comes from.

4 Working and Design Our project starts with the sensors collecting weather data in its surrounding environment and pushing it to firebase real-time-database and think to speak cloud. The data stored in firebase is used to display our React website and the data collected in

IoT Weather Forecasting Using Ridge Regression Model

737

Fig. 5 Model design

think speak is used to visualize and export the collected data as a CSV file which is later used to train our model using the Ridge Regression (RR) model. We have used the Ridge regression (RR) model to predict the weather. Ridge regression may be a model standardization methodology that’s accustomed to analyses any knowledge that suffers from multiple correlations. This method will use L2 regularization [12]. Uses the maximum temperature (Tmax) and minimum temperature (Tmin). Figure 5 describes the connection of sensors on a breadboard, where DHT 11, BMP 180, Rain sensor and ESP8266 Wi-Fi modules are connected.

5 Literature Survey Paper 1: Weather Prediction Using Machine Learning. Publication Date: 05 May 2021. Authors: Abhishek Patel, Pawan Kumar Singh and Shivam Tandon. Publisher: Galgotias University-Galgotias University School of Computing Science and Engineering.

738

K. G. Dath et al.

Abstract Climate conducts a completely critical function in many key production sectors, e.g., farming. Climate change with high charging these days, which is why old weather forecasts are getting closer and less powerful and continue to be annoying. Miles is therefore very important to decorate and modify the weather forecast model. those predictions affect the country’s financial system and people’s lives. A system of information and statistical analysis algorithms has been used that includes a wooded area used for weather forecasting [10]. Methodology This paper mainly focuses on the basic Machine Learning model, which has been an automated system to gather historical data. The paper focused to help future generations of people where prediction of weather helps lives to get prepared for the weather changes, taking precautions of bad weather conditions etc. Paper concentrated on format effective weather forecast and performing a function of accurately predicting the weather [10]. Paper 2: Smart Weather Forecasting Using Machine Learning. Publication Date: 25 August 2020. Authors: A H M Jakaria, Md Mosharaf Hossain and Mohammad Ashiqur Rahman. Publisher: A Case Study in Tennessee. Abstract Traditionally, weather predictions are performed with the help of large complex models of physics, which utilize different atmospheric conditions over a long period. These conditions are often unstable because of perturbations of the weather system, causing the models to provide inaccurate forecasts. The models are generally runon hundreds of nodes in a large High-Performance Computing (HPC) environment which consumes a large amount of energy. The paper presents a weather prediction technique that utilizes historical data from multiple weather stations to train simple machine learning models, which can provide usable forecasts about certain weather conditions for the near future within a very short period. The models can be run in much less resource-intensive environments [11]. Methodology This paper predicts the weather by collecting the historical data from multiple weather stations train simple machine learning models, which can provide usable forecasts about certain weather conditions for the near future within a very short period. The models can be run in much less resource-intensive environments [11]. Paper 3: Temperature Forecast Using Ridge Regression as Model Output Statistics. Publication Date: 30 April 2020.

IoT Weather Forecasting Using Ridge Regression Model

739

Authors: Niswatul Qona’ah, Kiki Ferawati, Muhammad Bayu Nirwana and Sutikno. Publisher: Universitas Sebelas Maret. Abstract This study uses the maximum temperature (Tmax) and minimum temperature (Tmin) observation at 4 stations in Indonesia as the response variables and Numeric Weather Prediction (NWP) as the predictor variable. The results show that the performance of the model based on Root Mean Square Error of Prediction (RMSEP) is considered to be good and intermediate. The RMSEP for Tmax in all stations is intermediate (0.9–1.2), Tmin in all stations is good (0.5–0.8) [12]. Methodology This uses the maximum temperature (Tmax) and minimum temperature (Tmin) observation at 4 stations in Indonesia as the response variables and NWP as the predictor variable. The result from Ridge Regression is more accurate than the Numerical Weather Prediction model and also it corrects up to 90.49% of the biased NWP for Tmax forecasting [12]. Paper 4: Device and Development of Automatic Microcontroller-based Weather Forecasting Device. Published Date: March 2020. Author: Dr Bindhu V. Publisher: PPG Institute of Technology, Tamil Nadu. Abstract The proposed method utilizes the sensors to monitor the weather changes and engages the raspberry pi to process the information gathered and convey it to the end user. The proposed system was tested by implementing it in the Indian delta districts and the accuracy, precision and flexibility in the forecasting were evinced by the data output observed over and done with the ThinkSpeak [13]. Methodology The proposed model develops a weather monitoring device to measure the meteorological parameters by employing the sensors, the sensors employed measure every minute changes in the atmosphere and convey it to the web server for the direct access of the users, proffering an accurate forecast of weather [13]. Paper 5: A Methodology of Atmospheric Deterioration Forecasting and Evaluation Yjrough Data Mining And Business Intelligence. Published Year: 2020. Author: J V Anand. Publisher: PACE Institute of Technology and Sciences, Ongole.

740

K. G. Dath et al.

Abstract This paper emphasis on an instinctual and efficacious method to forecast and analyze the condition of different atmospheric determinants all over the world. The difficulty in the extant remedies or the apparatus is its incapability to provide comprehensive information regarding the evaluation of the attributes of the atmosphere. The proposed methodology in the paper gathers the actual information about the atmospheric attributes such as the water, air, the forest and the tree cover etc. from the government bases and processes the collective information [14]. Methodology The methodology does the extrication transformation load over the original collective data that are in its raw format. The converted information sets are imported into the database to develop a dashboard with the multiple information displayed on it. This allows having evaluated data about the various atmospheric factors. To forecast the deteriorations and the conditions of the atmospheric attributes the methodology proffered utilizes the Fuzzy C means clustering, R-studio, and the ARIMA framework [14].

6 Equations n 

⎛ ⎝ yi −

i=1

p  j=1

⎞2 xi j β j ⎠ + λ

p 

β 2j

(1)

j=1

Ridge regression places a particular form of constraint on the parameters (β’s): β^ bridge is chosen the penalized sum of squares. which is equivalent to minimize   to minimization of i = 1n(yi − j = 1pxijβj) ^2 subject to, for some c > 0, j = 1pβj2 < c, i.e. constraining the sum of the squared coefficients.The related equation is shown in Eq. (1). Therefore, ridge regression puts further constraints on the parameters, βj’s, in the linear model. In this case, what we are doing is that instead of just minimizing the residual sum of squares we also have a penalty term on the β’s. This penalty term is λ (a pre-chosen constant) times the squared norm of the β vector. This means that if the βj’s take on large values, the optimization function is penalized. We would prefer to take smaller βj’s, or βj’s that are close to zero to drive the penalty term small [15].

7 Result Screenshots Figure 6 shows our implementation of the weather station using IoT components.

IoT Weather Forecasting Using Ridge Regression Model

741

Fig. 6 Model and connections of project

Figure 7 shows Arduino serial monitor which is displaying the messages which are written in the program. Figure 8 shows the Firebase Realtime Database, which is used as a backend for Reacts website. Figure 9 is the website which is disapplying of the weather information. Figure 10 shows the ThinkSpeak cloud which we have used to store the weather information. Figure 11 shows our data set. Figure 12 shows the graph of prediction from the data collected. The graph show the actual data and predicted data.

8 Conclusion Weather Forecasts is a project based on the Internet of Things that aims to give weather information and predictions using the Ridge Regression (RR) Model, a machine learning method. Ridge regression is a model tuning technique used to analyze data with multicollinearity. The challenges we faced while doing this project include collecting data from the sensors and storing it. Most of the tools used in this project are open-source tools. Rain is one of the data that we were unable to collect and the data collected is from constant locations instead of multiple locations. We aim at improving the project in future for better results.

742

Fig. 7 Serial monitor

Fig. 8 Realtime database

K. G. Dath et al.

IoT Weather Forecasting Using Ridge Regression Model

Fig. 9 Dashboard created using React JS

Fig. 10 Data collected and visualized by thing speak cloud

743

744

Fig. 11 Dataset used

Fig. 12 Predicted value with the actual value

9 Future Enhancement 1. Collecting weather data from multiple locations. 2. UV index.

K. G. Dath et al.

IoT Weather Forecasting Using Ridge Regression Model

745

3. Displaying predictions on the website. 4. Analyze wind speed and pattern.

References 1. The Arduino Team (n.d.) Getting started with Arduino products. https://www.arduino.cc/en/ Guide/ 2. React (2019) Getting started—React. https://reactjs.org/docs/gettingstarted.html 3. Google (n.d.) Firebase realtime database | firebase documentation. Retrieved 18 Mar 2022 from https://firebase.google.com/docs/database 4. https://realpython.com/jupyter-notebook-introduction/ 5. Nodemcu ESP8266. Components101 (n.d.) Retrieved 18 Mar 2022 from https://component s101.com/development-boards/nodemcu-esp8266-pinout-features-and-datasheet 6. Industries A (n.d.) BMP180 barometric pressure/temperature/altitude sensor-5V ready. Adafruit industries blog RSS. Retrieved 18 Mar 2022 from https://www.adafruit.com/product/ 1603 7. https://ae-bst.resource.bosch.com/media/_tech/media/product_flyer/BST-BMP180-FL000. pdf 8. Rain drops sensor module. Components101 (n.d.) Retrieved 18 Mar 2022 from https://compon ents101.com/sensors/rain-drop-sensor-module 9. DHT11 humidity and temperature sensor-Mouser.com (n.d.) Retrieved 18 Mar 2022 from https://www.mouser.com/datasheet/2/758/DHT11-Technical-Data-Sheet-Translated-Ver sion-1143054.pdf 10. Patel A, Singh PK, Tandon S (2021) Weather prediction using machine learning. SSRN: https:// ssrn.com/abstract=3836085 or https://doi.org/10.2139/ssrn.3836085 11. Jakaria AHM (2020) Smart weather forecasting using machine learning:a case study in Tennessee. arXiv.Org. https://arxiv.org/abs/2008.10789 12. http://sunankalijaga.org/prosiding/index.php/icse/article/view/533 13. Bindhu V (2020) Design and development of automatic micro controller based weather forecasting device. J Electron Inf 2(1):1–9 14. Anand JV (2020) A methodology of atmospheric deterioration forecasting and evaluation through data mining and business intelligence. J Ubiquit Comput Commun Technol (UCCT) 2(02):79–87 15. The Pennsulvania State University Ridge regression model equation. https://online.stat.psu. edu/stat857/node/155/

Automated Cloud Monitoring Solution: Review Ishwari Deshmukh and Jayshri D. Pagare

Abstract The Cloud adoption statistics clearly depicts that considerable number of organizations has hosted their infrastructure on cloud, most of the major migrations from on premise server to cloud are done and some are in progress. With this Cloud Computing era, Infrastructure as a service requires more efficient management. It comprises of IT resources for storage, networking, computing etc. Optimized and efficient use of the rented resources is the need of hour as resources are on rented basis and they have cost associated with it. As industry is growing, users are growing, resources are growing followed by volume and traffic. To manage any cloud infrastructure there is need of cloud monitoring solution with proactive alerting. This paper will provide entire overview with focus of providing open-source solution to monitor cloud environment. Keywords Cloud computing · Cloud monitoring · Monitoring phases · Beat agents · Alerting · Monitoring metrics · Parsers · Data shippers

1 Introduction Cloud Computing is a model that provides many on demand services. XaaS is the common terminology used for this purpose. It means that anything can be outsourced and integrated with existing application to improve performance or quality of service. As Cloud model comprised of three main services as Infrastructure as a Service, Platform’s a Service, Software as a Service. We also have technologies as Communication as a Service, Database as a Service, and Monitoring as a Service. To emphasis on Monitoring as a Service which is used to improve quality of Service and to attain service level agreements at accuracy. Apart from that it also helps to troubleshoot, find out root cause analysis, identify various vulnerability and threats over the application environment. I. Deshmukh (B) · J. D. Pagare Department of Computer Science and Engineering, JNEC, Aurangabad, Maharashtra, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_54

747

748

I. Deshmukh and J. D. Pagare

With each deployment model i.e. public cloud, private cloud, hybrid cloud there comes the need of monitoring the computing resources to use resources in optimised way. For public cloud resources are made available over the internet to multiple known or unknown users. Similarly private cloud exposes and provides access to limited and most of the known audience. However, hybrid cloud is combination of above-mentioned cloud, but monitoring aspects remain constant irrespective of deployment model. Scaling up of resources or scaling down of resources plays an important part. Organization often terms it as capacity management or capacity planning. Entire moto of the proposed system its t monitors the environment with least cost and best quality of service with minimal or no failure of computing resources.

2 Motivation to Design With increasing complex cloud infrastructure, it is especially important to have a stable monitoring solution applied over the cloud cluster. Moreover, existing resources are on rented basis i.e., already cost is to be paid for them. Resources here refers to the processors, storage, or network [1]. There is a need to identify and design an architecture that is efficient to monitor the cloud infrastructure with minimal or no cost. This solution should be effective and reliable as very less of negligible cost is associated to build and maintain it [2]. Cloud Computing involves many activities for which monitoring is an essential task. The most important ones are like. (1) (2) (3) (4) (5) (6)

Capacity and Resource Planning and Management. Security management. Data Center Management. SLA Management. Billing. Performance Management.

Traditional monitoring or availed monitoring services are mostly domain specific in nature. This is an attempt to provide a versatile, low cost and effective solution for cloud infrastructure [3]. Using Monitoring as a service from cloud vendor will ultimately require cost of two thing. Firstly, it will be all the rented resources or platform, or it can be software service that is used directly [4]. Second part will comprise of Monitoring service which will continuously run over the cloud infrastructure, and it will cost per minute [1, 5]. In Order to provide cost effective open-source solution that can be designed and used as per domain specific need one can use various automation tools combine them and create a complete open-source solution. Hence Monitoring has become an important aspect that is to be considered while designing as well as maintaining cloud infrastructures. Monitoring the cloud infrastructure and analysing the gathered statistics helps for capacity management. Resources or instances can provision based on the need and as required they can be decommissioned. This ultimately defines on demand pay as

Automated Cloud Monitoring Solution: Review

749

used model [6]. Automation helps to upscale and downscale the system. No human intervention is required in entire process and the accuracy is high. This defines the term rapid elasticity which the monitoring solution will provide. Used services can be measured, monitored, and reported [5].

3 Architecture of Monitoring System Monitoring architecture can be broadly classified in three major steps. (1) Data Collection (2) Data Analysis (3) Data Visualization. These three steps show high level picture of any monitoring solution. From granular perspective there are multiple operations that happens within this phase. These included collection of data, transportation of data, processing the data at required processing time, presenting thee data in most effective way [1, 7]. Apart from that every phase must be associated with some security protocol so that security breach can be avoided. Cloud monitoring architecture is shown in Fig. 1. Basic requirement to set up any monitoring solution is environment that is to be monitored, light weight shippers to ship data, on demand parser, storage for gathering the stats, UI to visualize the gathered data. These are basic primary requirement for any monitoring solution that is to be built [8]. Initial stage design of architecture plays a significant role. Infrastructure to be monitored need to be measured as all the instances will forward the statistic to main centralized server. Average no of hits per minute need to check in order to build the centralized server [1, 5, 7]. One can design it as master server architecture, where the master comprises of universally available and efficient parser and slaves, or the servers are the devices that needs to be monitored [9].

Fig. 1 Cloud monitoring architecture

750

I. Deshmukh and J. D. Pagare

Phase one comprises of data collection. Here data can be referred as files, log files, metrics, information stats of cloud services, health rules for the instances, etc. [3]. Broadly this monitoring solution can be classified into two types based on the data. It can Application monitoring or Infrastructure monitoring. Application monitoring is collection of application logs, application of grok patterns on it and storage in form of attribute and value. E.g.: A java application has various exceptions. One can setup alert if any such exception string occurs in log. Infrastructure monitoring is where system related metrics are gathered e.g.: CPU, Memory, Disk, Process, etc. Alerts can be added if any of the mentioned parameter breaches the threshold value that are defined already [10]. Overall, any event that is to be captured and monitored or visualized can be considered as data set. Although it depends upon design architecture what data to collect and what part of data to be extracted. To gather this data once can use beat agents [1, 11]. These agents are lightweight programs that run on the server. These shippers run on server as service, and they run throughout in the background without consuming resources. Main function of this agent is to push data to centralized server or the parser. In some architecture where parser is absent it could be direct storage. It can be static database of cloud availed database service. One can gather events from automation scripts. Also, various API can be used to fetch data. Collection of data can also be done with plugins or protocols such as Nagios or NRPE. Ready to install beat agents are available which are mainly written in java or GO language one can ship data without constraint of compatibility [2, 6, 7, 12]. Transportation of events or logs are mainly taken care by the shippers. It’s important to have encrypted mechanism if data is shipped with hops over the cluster. This will help to avoid all the security breaches [13]. Phase two comprises of as centralized server where all the data is accumulated and sent over to storage device or storage service. Each and every event that gets generated with help of beat agent is undergone through a parser device. Here various grok patterns can be placed. Segregation of data can be done at this level. Key value pair format can be used to store data. This parsing helps to visualize the data on graphical UI easily. One can plot various visualizations on segregated data and analyse the behaviour [2, 14]. The centralized server designed in this phase should be capable to oversee all the network load. Network load here refers to the events that are transferred all the way from cloud infrastructure. Parser or number of parsers should be defined based on no of events that we are going to receive. One can receive any number of events with any frequency. Mostly frequency started from 2 s. However, this is relative to the use case, if application is critical one can reduce the time to seconds otherwise its ca be in minutes interval [15]. Based on that infrastructure should be designed. Parser should be designed in such a way that it should oversee all the load. Multiple nodes of the parser can be installed. Nodes should be available with HA mode i.e., high availability mode. If one of the instances goes down other should be capable enough to handle all the traffic in the environment. Also, for larger clusters load balanced parsers can be applied. Events load will be routed and distributed evenly over the load balance parser [2, 16].

Automated Cloud Monitoring Solution: Review

751

Phase three mainly focus to store and visualize the data. As mentioned earlier storage device should be efficient in terms of volume. It can be static storage or simple database or cloud service as s3. Elasticsearch database can be used here to store the data in Json format. One data get segregated with key value pair we can use it easily to analyse the data behaviour [1, 2]. Other important aspect is to hire the storage that can oversee substantial number of read write operation within low frequency and should provide efficient output with near real time. Time lapse parameter should be considered as further alerting or visualizing process is dependent on this. Data can be aggregated and stores in attribute value format. This can be depicted in form of graph, line chart, pie chart, heap map, etc. Various dashboards and canvas reports can be made out of it [17]. Stored data can be aggregated on any configured attribute that is present in logs which makes fault finding and debugging easy. Alerts can be configured on stored data in form of email. Also, integration can be extended to any destination or any ticketing tool for further analysis. Also, Machine learning modules can be applied over the data. Module can predict number of resources required based on prior knowledge also they can predict the failure in applications [1, 2, 18].

4 Results and Performance Analysis This architecture can be built in both ways, i.e., centralized or decentralized as in Fig. 2. A solution can have hardware device with CPU that has processor greater than two cores with Memory more than 4 GB. Any Linux flavours will suffice the need. Light weight agent is installed on the system. Agent is java-based agent that is built in GO lang. Agent varies as per need. E.g. If system statistics are to be gathered then metric beat type of agent to be installed on the system. If file data or log files need to be stored, parsed, and monitored then file beat type of gent is required to be installed. Similarly, if application health or URL are to be monitored then heartbeat type of agent to be installed. Events that are getting collected can be parsed. The required strings are to be captured and stored in database. Unusual data can be discarded at initial stage. Parser collects the data at specific port e.g., 5044. Once event is received segregation of string happens so as to store the data in attribute value format. Time stamp data or real time data is parsed. Configuration file can be considered of three important parts namely input, filter, output. Input part listen to mentioned port. Filter part segregates the data and directs it to desired index mentioned in output part. Here the output is Elasticsearch type of database where data is stored. All graphical interfaces run on this. Gathered data is seen in form of indices. Indices are used to create aggregated Dashboards. Canvas reporting the real time reporting is built on top of it. Java program watchers can be used to monitor the data. Every 5 min a program gets executed to identify if all nodes are reporting and all the nodes in the environment are performing well. Integration with SMTP server is required if Elasticsearch data alerts are to be sent

752

I. Deshmukh and J. D. Pagare

Fig. 2 Designed architecture

in email format. This gives complete pictorial view of application and infrastructure on single dashboard. Replication of proposed solution is easy as time taken for development is one time effort. Solution being scalable in nature can be scaled using various automation tool. Agents or monitoring shippers can be installed across entire infrastructure in using single playbook. Expansion of parser requires replication of existing parser. To generalize the fact, time complexity to expand the solution is minimal.

5 Advancement As we have pass dated information that is parsed and stored over the cloud which is nothing but the data set for further uses. Here we can use the data set in two ways as prediction and dynamic alert configuration. Prediction comprises of analysing the data. Identifying the trends for past dates and predicting the future based on past data. This can be used for early resource planning. For example, streaming platform where traffic gets increased on holidays. We can refer to the trends that are predicted by the machine learning model to plan server capacity that can fulfil request without delay. Similarly, this can be used for applications business management [4]. With the same predictions, adjustment of thresholds can be done in dynamic manner. Alerting part disused in phase three above has threshold values that were define. These thresholds are generalized as certain number of resources that should be consumed for application to be active. However, some positive or negative deviation in this consumption is sign of some discrepancy. Dynamic threshold definition will help for proactive monitoring [19].

Automated Cloud Monitoring Solution: Review

753

6 Comparison of Cloud Monitoring Solutions In comparison to existing solution, we have below mentioned highlights that make the system more effective. Here existing solution mainly refers to the on-demand services provided by cloud service providers as Amazon, Google or Microsoft. It can be also any monitoring tool as Nagios or AppDynamics. Tool Nagios uses NRPE kind of protocol that acts as beat agent to transfer data from source to destination. Nagios allows users to design and deploy customised plugin over the infrastructure. This designing of plugins is easy and ca be used. AppDynamics is has generic architecture wherein shippers are associated to transfer data. However, dynamic real time monitoring can be obtained using it. In comparison to all these tools given solution has below mentioned benefits [8, 11]. Cost Effective: To use any of the service one has to pay the cost with pay per use module. For monitoring service, the use is very high as infrastructure or application should be always under monitoring to avoid outages. Above given solution is totally open source and no cost is required and hence monitoring is free. However minimal storage cost is required if data storage to be done. Customized Monitoring: One can monitor application logs and system logs. Grok pattern can be applied on data to find error, warnings or any of the desired string. One can parse and find various error codes. Application end points can be polled at certain interval to check if they are up. This all leads to proactive monitoring and zero downtime. Easy to Setup and Maintain: Installation of agents over the servers is simple job with automation tools as ansible or chef. Single server with playbook can install agents on multiple servers within less time. Similarly, maintenance and upgrade become one step activity. Scalable Solution: This solution can be implemented on any of the cloud infrastructure. Scaling of application is quite easy as one has to replicated the parser among various load balancers if required. Also, storage can be scaled if traffic is high. However, with all the changes implementation logic remains same. Static/Dynamic Proactive Alerting: As mentioned earlier, alerting works on static thresholds as well as it works with dynamic threshold. Alerting being proactive i.e., alerting user before issue occurs. This will provide minimum or no outage over the application business [1, 4].

7 Conclusion Client machines are the small nodes that will be idle but treated as application servers. These machines will have shippers that are placed over. These shippers will be lightweight and can be installed manually or in automated way as well. Small

754

I. Deshmukh and J. D. Pagare

programs will be running all the time to monitor the instance. Solution being open source will be easy to manage and maintain in cloud infrastructure. Now we have multiple client that will reside in the cloud with that we will have centralized server that will be in same environment. Metrics will be fetched from clients and can be stored on server. Analysis can be done on saved data. Data can be visualized into gauge, graphs, line charts, bar charts. This dashboard can be plotted as and when required. Moreover, alerting, and proactive alerting can be set up based on data requirements. This alerting feature will help the user to get alerts when something is going wrong or few of the features in application are not working as expected. This will help end user to identify and fix the failure before system gets crashed. One can analyses entire cluster in once screen. This solution is scalable and can be replicated over any cluster. This solution being open source can be replicated. It can be placed on two nodes clusters to two thousand with infra design that is capable to handle the load.

References 1. Birje MN, Bulla C (2019) Cloud monitoring system: a review. Int J Eng Sci Manage 1(1):49–55. A Multidisciplinary Publication of VTU 2. Sun Y, Xiao Z, Bao D, Zhao J (2019) An architecture model of management and monitoring on cloud services resources. In: 3rd International conference on advanced computer theory and engineering, ICACTE, vol 3. IEEE, pp V3–207–211 3. Ali A (2022) An overview of cloud computing for the advancement of the e-learning process. J Theor Appl Inf Technol 100(3) 4. Bulla C, Bhojannavar S, Danawade V (2013) Cloud computing: research activities and challenges. Int J Emerg Trends Technol Comput Sci 5. Alwadan T (2018) Cloud computing and multi-agent system: monitoring and services. J Theor Appl Inf Technol 96(09) 6. Meera A, Swamynathan S (2013) Agent based resource monitoring system in IaaS cloud environment. In: International conference on computational intelligence: modelling techniques and applications (CIMTA) 7. Ward JS, Barker A (2014) Observing the clouds: a survey and taxonomy of cloud monitoring. J Cloud Comput Adv Syst Appl. Springer 8. Cloud Computing (2019) National Institute of Standard and Technology, 05 Feb 2019. https:// nvlpubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication800-145.pdf 9. Aversa R, Tasquier L, Venticinque S (2017) Agents based monitoring of heterogeneous cloud infrastructures. In: 10th International conference on ubiquitous intelligence and computing and 2013 IEEE 10th international conference on autonomic and trusted computing. IEEE 10. Zubac DA, Kravchenko TV, Maiatin AV, Khagan MV (2018) A multi-agent approach to the monitoring of cloud computing system with dynamically changing configuration. In: FRUCTISPIT. IEEE 11. Birje MN, Bulla C (2019) Cloud monitoring system: basics, phases and challenges. Int J Recent Technol Eng (IJRTE) 8(3). ISSN: 2277-3878 12. Zhang C, He Y, Du B, Yuan L, Li B, Jiang S (2020) Transformer fault diagnosis method using IOT based monitoring system and ensemble machine learning. Futur Gener Comput Syst 108:533–545

Automated Cloud Monitoring Solution: Review

755

13. Barot V, Kapadia V, Pandya S (2020) QoS enabled IOT based low-cost air quality monitoring system with power consumption optimization. Cybern Inf Technol 20(2):122–140 14. Alsubari SN, Deshmukh SN, Alqarni AA et al (2022) Data analytics for the identification of fake reviews using supervised learning. Comput Mater Continua 70(2):3189–3204 15. AlKhunzain A, Khan R (2021) The use of M-learning: a perspective of learners’ perceptions on M-blackboard learn 16. Hu S (2022) Research on monitoring system of daily statistical indexes through big data. Recent Adv Comput Sci Commun 15(5):731–740 17. Klaver L et al. (2021) Towards independent runtime cloud monitoring. In: Proceedings of 2021 ACM/SPEC international conference on performance engineering (ICPE’21), Virtual Event, France, 19–23 Apr 2021. ACM, New York 18. Toczé K, Madon M, Garcia M, Lago P (2022) The dark side of cloud and edge computing: replication package. https://doi.org/10.5281/zenodo.6390610 19. Khalil MM et al. (2022) Measuring the effect of monitoring on a cloud computing system by estimating the delay time of requests. J King Saud Univ Comput Inf Sci 34(7)

A Secured Framework Against DDoS Attack in Wireless Networks O. K. Vismaya, Ajay Kumar, Arya Paul, and Albins Paul

Abstract In the modern world, wireless networks are more prominent. Due to the dynamically changing nature, mobile networks are more vulnerable to several attacks. Data security is a significant problem since data flow across a wireless channel that is naturally exposed, making possible for malicious attackers to obtain sensitive data. Finding and resolving security problems in wireless networks are essential steps in ensuring secure data transmission. One of the main threats that can be performed against wireless networks to stop legitimate nodes from sending and receiving data packets is distributed denial of service (DDoS). The proposed framework includes implementation, detection and elimination of DDoS attacks in the wireless network. To provide further security, an authentication approach is introduced which will grand access/denial to the receiver based on successful authentication of the nodes. Finally, analysis of the proposed model is performed in 3 modes i.e., normal mode, attacker mode and controlled mode. Keywords Distributed denial of service · Wireless networks · Authentication · Ad hoc on demand distance vector routing

O. K. Vismaya (B) · A. Kumar · A. Paul · A. Paul Adi Shankara Institute of Engineering and Technology, Vidya Bharati Nagar, Ernakulam, Kerala, India e-mail: [email protected] A. Kumar e-mail: [email protected] A. Paul e-mail: [email protected] A. Paul e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_55

757

758

O. K. Vismaya et al.

1 Introduction Today, the internet has evolved into a need for everyday life and is utilized for more than just amusement. It also facilitates ordinary tasks like money transfers, bill payments, reservation of tickets, educational analysis, viewpoints on learning, commerce, media coverage, etc. There are primarily two fundamental varieties of networks: wired networks and wireless networks, and the main goal of computer networks is to exchange resources. In computer networks, nodes will make a connection with the use of cables like fiber-optic, twisted pair, and coaxial, to exchange data through a connection known as a data link. A wireless network (WLAN) allows access to the network without a cable connection because the network is set up via frequency signals. It refers to using a certain domain (range), such as WIFI, to access internet services without requiring a physical connection (stands for wireless fidelity). Desktops, mobile phones, and servers are examples of nodes or hosts in computer networks. Each has a special code known as a MAC address. Early on, diversity develops as a result of the market sales of switches, routers, and other equipment by many network manufacturers [1]. Depending on the need, a variety of wireless systems including WLAN, WPAN, WMN, and WSN are available. However, there is a significant security problem with wireless networks, particularly with regard to specific attacks that depend on the medium and are not present in the older counterpart. Ping-pong effect, MITM, exhaustion, collision, radio jamming, signal jamming, noise jamming, unfairness, PAN id conflict, capture, and tempering attacks are examples of traditional DDoS attacks that are conducted in the PHY and MAC layers[2]. The wireless media introduces a number of attacks that are difficult for standard protection techniques to effectively counter. Distributed denial-of-service (DDoS) assaults are a prominent class of these attacks that target system or user domain buffers. These attacks launched against wireless networks and will prevent legitimate nodes from transmitting and receiving. However, in the wireless world, attackers might be able to seize the communication channel and stop genuine nodes from communicating. Because wireless networks are built using an open medium, intruders have an easy way to implement such attacks. A straightforward DDoS assault can defeat wireless network defenses like cryptography and pass-phrase sharing, effectively shutting down the entire network. Thus, in order to prevent the effect of attacks, an efficient counter measurement system must be developed and make the network more secure[3, 4].

2 Related Works Poongodi et al. [5] proposed a method to effectively protect data against a selective drop assault, this study provides a resistant to selective drop attack (RSDA) technique. The neighboring nodes won’t faithfully pass the information to the neighboring nodes during a selective drop attack. A rogue node, however, that has placed in a

A Secured Framework Against DDoS Attack in Wireless Networks

759

data transmission route, has the ability to block certain forwarding messages. Malicious nodes that are overtaxing a host and rendering it completely unusable must be found. The hostile nodes would refuse to forward messages travelling through them in a selective drop attack. Finally, the assault reduced the host’s throughput to its minimal value. Shivam Dhuria and Sachdeva [6] introduced a model to protect Wireless Sensor Networks in an efficient way. In this work, two approaches are presented. The first is a light-weight two-way authentication method that will shield WSNs from the majority of attacks, and the second is a traffic analysis-based data filtering method that will identify and shield WSNs from DDoS attacks. Several performance metrics have been used by the Network Simulator 2 (NS2) to verify the findings. i.e. throughput, delay, lost packets, energy consumption and PDR. Qazi et al. [7, 8] proposed a method to secure wireless Sensor Networks (WSNs). This study primarily targeted security challenges in WSNs. Using Elliptic Curve Digital Signature (ECDSA) algorithm, the introduced model not only secures the communication from one node to the other node network but also stores space of memory in one node to provide the correct method for calculating the time for generating key, number of hello message, and size of the packet. Additionally, key management with suitable key length is offered by the Algorithm for Wireless Secure Communication (ASCW). Additionally, ASCW aids in safeguarding node-level communication, aiding greater and more effective network security. With the aid of the authentication process, ASCW also lowers the cost of risk and security concerns on the network. Kshirsagar et al. [9] proposed a model for mobile ad hoc network (MANET). It is a mobile node communication network that lacks any preceding communication infrastructure[10]. The suggested technique is compatible with the current routing protocol. Secure communication paths use the trusted list idea. The trusted list and trust values display the number of times each node engaged in communication. The energy level of mobile components is used in MANET to distinguish between altruism and selfishness. Security and the distinction between selfishness and altruism are determined using the trust and energy models, respectively. Kolandaisamy et al. [11] introduced a method for detecting Distributed denial of service attack (DDoS). The approach determines the trustworthiness of the packet using all these variables and takes it into consideration while making decisions [12–14]. The proposed SPPA model bases its determination of the detection of a DDoS assault on the CCA’s estimated value. Every single vehicle’s behavior is evaluated, and the true weight of the car or node is determined. Every single vehicle’s behavior is evaluated, and the true weight of the car or node is determined. The assault detection is carried out using this valid weight by determining whether the vehicle is an intruder or a regular node. Amrish et al. [15] proposed a solution to determine DDoS attacks. Cyberattacks of the Distributed Denial of Service (DDoS) variety aim to overload a target server in an effort to stop regular traffic. While being subjected to a DDoS assault, the system is still busy processing requests from bots rather than serving actual users. In this work, normal traffic and DDoS attack traffic are separated using a machine learning method. There are four machine learning classification approaches used to find DDoS attacks. The Artificial Neural Network (ANN) produces the greatest results when measured against KNN, Decision Tree, and Random Forest. Smys [16]

760

O. K. Vismaya et al.

introduces a model to detect DDoS attck in the communication network, which is made up of the terminal nodes. The telecommunications networks are vulnerable to a variety of cyberthreats, with distributed denial of service (DDOS) being the one that affects customers’ access to services the most frequently. Therefore, the study offers a detection and classification approach for DDOS attacks in the communication system using a neural network and support vector machine combination. Performance evaluation of the proposed model is performed using NS2.

3 Proposed Model A number of computer users have been increased dramatically and tremendously along with the interest in internet usage. The rapid increase of laptop computers and PDAs has increased the locations where individuals undertake computing tasks, including schools, colleges, business centres, and even homes. Everyone wants to join wireless networks because they provide users with mobility. While considering this type of network security, the message is a primary concern because there are many users in the network. The wireless networks are vulnerable to various attacks that can harm the network. Security of messages is a primary concern because there are a large number of users. Distributed denial of service is one of the main attacks in the network which can disrupt the services provided by the network. It is an active type of attack, which have the ability to make changes to the data while routing. The DDoS attack could be considered as one of the main threats to mobile networks. To provide security to the network, a secured framework is introduced in the wireless networks against DDoS attack. Block diagram for the proposed model is shown in Fig. 1. In this proposed model DDoS attack will be detected and will get eliminated. In addition to eliminating attack from the network, an authentication mechanism is introduced which will authenticate the data and will ensure more security to the network. After authenticating the nodes, the data will be routed using the AODV protocol. Thus, as a whole the proposed model is divided in to four phases they are: • • • •

Phase I: DDoS attack implementation Phase II: Attack detection Phase III: Attack elimination Phase IV: Authentication.

Fig. 1 Block diagram of proposed model

A Secured Framework Against DDoS Attack in Wireless Networks

761

3.1 Phase I: DDoS Attack Implementation The most well-known assaults used to bring down the entire network are DoS. By delivering malicious requests, DoS or DDoS attacks can fully deplete the network’s resources. It is also a form of active attack that affects the network’s availability. DDoS refers to DoS that can be performed by several nodes, increasing the number of network enemies. The attack model is employed in the wireless network to better understand the impacts of DDoS assault on a network. Initially, an attack model is implemented over the normal ad hoc on-demand distance vector. For the implementation of the first phase, 20 normal nodes and 5 attacker nodes are defined from the total of 25 nodes.

3.2 Phase II: Attack Detection DoS attack detection is performed in the phase II. This detection is performed based on the rate of packets that are dropped from the network during data transmission. For each packet drop, route flag will become 1, which indicates the packet drop. At the same time, an initial trust value of zero is kept for each node. This trust value will get incremented for every positive flag 1 and the node which having the trust value above a particular threshold value will be considered as an attacker node. All the identified attacker nodes will be then saved to a text file.

3.3 Phase III: Attack Elimination After detection of attack from the network, the detected attacker nodes will be eliminated from the network which is performed in the phase III. In this phase, initially the system will read the text files by which attacker nodes are saved. Then the system will exclude those attacker nodes from the network. In addition to that, there may be a chance of including the attacker nodes by default to the routing table. In such a case the system will again check the presence of attacker nodes from the routing table and will eliminate those identified attacker nodes from the network. In such a way the attacker nodes can be eliminated effectively from the network.

3.4 Phase IV: Authentication To make the network more secure after the elimination of attacks from the network an authentication step is also introduced, which is considered phase IV. Proposed model for this authentication phase is shown in Fig. 2. In this phase, it includes both

762

O. K. Vismaya et al.

sender and a receiver. The sender will send an authentication message code. Then the message code will be converted to hash code using rotational hashing. After that a private key X will be generated using Elliptical Curve Digital Signature Algorithm (ECDSA). ECDSA will be using a sign digital signature value, they are two values r and s. Each node will be having a public and private key. The private key used in this model is X and public key is the combination of 3 values they are P, G and Y. P is a randomly generated prime number, G is a constant and Y is a value that is generated by using the private key. The public key Y is generated by finding the power of g and x. An authentication key is generated by using the public key and previously generated hash code. Then, using ECDSA algorithm generate the signature S and R. Finally compute T 1, T 2, and T 3 using the digital signature, private key and hash code. The receiver will then reconstruct the private key generated by the sender using the public key variables P, G and Y. Then compute TT1, TT2, ans TT3 using the signature S and R. Finally reconstruct the key and hash code. Finally compare the receiver side code (TT3) with the sender code (T3). If both the value is equal, then the authentication will be successful and the nodes will be permitted for communication otherwise the nodes will deny the communication. All the nodes in the network will participate in the communication.

Sender

Authentication code converted to hash code

Authentication successful

Yes

Private key generation

Authentication code generation using generated private key

Code received No

Receiver

Nodes permitted for communication

Authentication failed

Nodes denied communication

Fig. 2 Proposed secured framework

Public key generation using private key

Authentication key generation using private key and hash code

Reconstruct private key using public key

A Secured Framework Against DDoS Attack in Wireless Networks

763

4 Result Analysis In the proposed model DDOS attack is implemented in the wireless network. The implementation is performed in Network Simulator platform. There are certain parameters used for the implementation of attack and they are shown in table 1. Network Simulator simulates the proposed algorithm (NS-2). An open-source network simulation tool is called NS-2. ˙It will provide sufficient support for the simulation of different protocols in the wired and wireless networks. It will also provide sufficient support for different networking protocols, elements and routing types. The proposed design topology uses 20 normal nodes and 5 attacker nodes are present. The entire mechanism makes use of 25 number of nodes. AODV is the protocol in use. To compare the findings, different parameters are taken into account. In NS-2, the 802.11 MAC layer is implemented for simulation. Link uses a wireless channel type with an omni-directional antenna type. The MAC type used in this simulation is 802.11. The simulation time is 10 s, and the routing protocol is the AODV protocol. For simulation, the two-ray ground propagation model is employed. Priority queueing is the sort of interface queue that is employed. First phase of this model is to implement distributed denial of service attacker nodes in the network. the red colour nodes indicate the attacker nodes and the green colour nodes indicate the normal nodes. The attacker nodes are implemented continuously by sending repeated route requests to the network and make the node malicious. From the Fig. 3 as shown the nodes 4, 5, 14, 17, and 24 is set as the attacker nodes and the remaining nodes act as normal nodes. The dotted lines show the transmission of data during that particular time period and the circle shows the network coverage of that particular nodes. The detected attacker nodes are displayed in the output terminal. The attacker nodes which cross the threshold value will be detected as an attacker. Then the detected attacker nodes will be excluded from the network. Moreover, there is a probability of adding the attacker nodes to the routing table due to shortest path Table 1 Simulation parameters

Parameters

Value

Simulator

NS2 (Ver 2.35)

Normal nodes

20

Routing protocol

AODV

MAC

IEEE 802.11

Simulation time

10 s

Channel type

Wireless channel

Propagation

Two ray ground

Antenna type

Omnidirectional

Interface Queue Type

Priority queuing

No. of routing packets

12,000

764

O. K. Vismaya et al.

Fig. 3 Implementation of attacker nodes

finding property of AODV protocol. Due to this reason, the system will check the routing table again to verify whether any attacker nodes are added. Finally, the identified attacker nodes from the last iteration will be eliminated from the network. The hash code, private key, public key and the authentication key will be displayed in the output terminal. The hashing used in this proposed model is rotational hashing. The system will generate different hashing codes for each step and ensure more security to the network. Each node will be having its own private and public keys. The public key can be shared among others while the private key should be kept secretly. To analyze proper functioning of the proposed model certain parameters like packet loss, throughput, energy and end to end delay is evaluated: • • • •

Total packet drop Total Throughput Average energy End to End delay.

The attacker mode will drop a large number of packets when compared to the attacker mode. Throughput is the maximum amount of successful data packet that can be received in the destination. Thus, the attacker throughput will be low as compared to the normal mode and controlled mode. Similarly, a large amount of energy will be consumed by the attacker node when compared to the normal and controlled mode. The end to end delay also will be less for the secured framework. The performance analysis is shown in Table 2. Throughput performance of the system with normal, attacker mode and controlled mode is shown in the Fig. 4. In the normal mode i.e. without attacker, max number of packets will be received in the destination. In the attacker mode, the number of packets received will be less. While in the controlled mode the secured framework

A Secured Framework Against DDoS Attack in Wireless Networks

765

Table 2 Performance analysis S. No.

Parameters

Normal mode

Attacker mode

Controlled mode

1

Total packet drop

9397

10,110

5383

2

Total Throughput

2069

1297

2280

3

Average energy

1.89 J

10.04 J

1.54 J

4

End to end delay

0.38 s

0.63 s

0.19 s

will increase the rate of successful packet delivery. From the Fig. 4, it is clear that the designed framework will give a better performance compared with the other 2 modes. x-axis is plotted as time and y-axis is the number of throughput. From this it is clear that the system performance is low in the attacker mode. Performance of the system in the normal, attacker and controlled mode based on the packet drop is shown in the Fig. 5. Packet drop will be more in the attacker mode. While the drop in the normal and contolled mode will be less compared to the attacker node. The system will detect the attacker node when performance of the system misbehaves from the normal mode. Average delay for the normal, attacker and controlled mode is shown in Fig. 6. As the effect of attacks in the framework increases, the delay will also increases. Thus from Fig. 6, it is clear that the attacker effects in the framework is more copared to the normal and the controlled mode.

Throughput (kbps)

Throughput 3000 2500 2000 1500 1000 500 0 1

2

3

4

5

6

7

8

9

10

Time (seconds) Normal mode

AƩacker mode

Fig. 4 Throughput graph for normal, attacker and controlled mode

Controlled mode

766

O. K. Vismaya et al.

Packet drop (kbps)

Packet drop 15000 10000 5000 0 1

2

3

4

5

6

7

8

9

10

Time (seconds) Normal mode

AƩacker mode

Controlled mode

Average delay (seconds)

Fig. 5 Packet drop graph for normal, attacker and controlled mode

Average delay 0.8 0.6 0.4 0.2 0 1

2

3

4

5

6

7

8

9

10

Time (seconds) Normal mode

Aacker mode

Controlled mode

Fig. 6 Average delay graph for normal, attacker and controlled mode

5 Conclusion Widespread deployments of numerous wireless networks of varying sizes, including wireless personal area networks (WPANs), local area networks (WLANs), metropolitan area networks (WMANs), and wide area networks, demonstrate the rapid progress that wireless networking has experienced (WWANs). These wireless networks can have many organizational structures, including cellular, ad hoc, and mesh networks. They can also be specialized networks, such sensor and vehicle communication networks. However, because the underlying communications are carried out via electromagnetic radiations in open space, wireless networks lack physical security. For those who work in computer and network security, wireless networks present a special difficulty. Data that is transmitted through a network during communication is never secure since various attackers have access to the data. The data are encrypted to transform them into an unreadable form since the data must be secured from the attack.

A Secured Framework Against DDoS Attack in Wireless Networks

767

In future, this work can be extended from wireless networks to wireless sensor networks. The designed attacker topology can be implemented in other trust-based frameworks and can analyse the performance. Similarly, the designed secured framework can be implemented for multiple attacks.

References 1. Nazir R, Kumar K, David S, Ali M (2021) Survey on wireless network. Arch Comput Methods Eng 1–20 2. Bendale SP, Prasad JR (2018) Security threats and challenges in future mobile wireless networks. In: 2018 IEEE global conference on wireless computing and networking (GCWCN), pp 146–150 3. Balarengadurai C, Saraswathi S (2018) Comparative analysis of detection of DDoS attacks in IEEE 802.15. 4 low-rate wireless personal area network. Proc Eng 38:3855–3863 4. Sharma S, Mishra R, Singh K (2017) A review on wireless network security. ˙In: International conference on heterogeneous networking for quality, reliability, security and robustness, pp 668–681 5. Poongodi T, Khan MS, Patan R, Gandomi AH, Balusamy B (2019) Robust defense scheme against selective drop attack in wireless ad hoc networks. IEEE Access 7:18409–18419 6. Dhuria S, Sachdeva M (2018) Detection and prevention of DDoS attacks in wireless sensor networks. In: Networking communication and data knowledge engineering, pp 3–13 7. Qazi R, Qureshi KN, Bashir F, Islam NU, Iqbal S, Arshad A (2021) Security protocol using elliptic curve cryptography algorithm for wireless sensor networks. J Ambient Intell Humanized Comput 12(1):547–566 (2021) 8. Ali S, Humaria A (2020) An efficient cryptographic technique using modified Diffie–Hellman in wireless sensor networks. Int J Distrib Sens Netw 16(6):1550147720925772 9. Kshirsagar VH, Kanthe AM, Simunic D (2018) Trust based detection and elimination of packet drop attack in the mobile ad-hoc networks. Wireless Pers Commun 100(2):311–320 10. Cetinkaya A, Ishii H, Hayakawa T (2019) An overview on denial-of-service attacks in control systems: attack models and security analyses. Entropy 21(2):210 (2019) 11. Kolandaisamy R, Noor RM, Kolandaisamy I, Ahmedy IB, Kiah ML, Tamil EM, Nandy T (2021) A stream position performance analysis model based on DDoS attack detection for cluster-based routing in VANET. J Ambient Intell Humanized Comput 1–14 12. Riya soni1, Rajneesh Pachouri, Anurag Jain (2021) DDoS attack detection and prevention on wireless sensor network by using TBT method. Int Res J Eng Technol. p-ISSN: 2395-0072 13. Ghannam R, Sharevski F, Chung A (2018) User-targeted denial-of-service attacks in LTE mobile networks. In: 2018 14th International conference on wireless and mobile computing, networking and communications (WiMob), pp 1–8. IEEE 14. Bashirpour H, Bashirpour S, Shamshirband S, Chronopoulos AT (2018) An improved digital signature protocol to multi-user broadcast authentication based on elliptic curve cryptography in wireless sensor networks (WSNs). Math Computat Appl 23(2):17 15. Amrish R, Bavapriyan K, Gopinaath V, Jawahar A, Kumar CV (2022) DDoS detection using machine learning techniques. J IoT Soc Mob Analytics Cloud 4(1):24–32 16. Smys S (2019) DDOS attack detection in telecommunication network using machine learning. J Ubiquit Comput Commun Technol (UCCT) 1(01):33–44

Anomaly Based Intrusion Detection System Using Rule Based Genetic Algorithm Shraddha R. Khonde

Abstract In emergent field of networks everyone is able to access data as required. Huge amount of data transmission is done on internet; so data security, integrity and confidentiality become important. Data Security is improved by use of intrusion detection system (IDS).This system allows administrator to monitor network to keep it secure from vulnerabilities. Most intrusions happen in the network are due to an attack. A model based on hybrid structure for intrusion detection system using rule based genetic algorithm for anomaly detection for new identified attacks is presented here. Model makes use of various algorithms of machine learning such as naïve bayes, support vector machine and random forest into ensemble. Ensemble approach helps in performance improvement. Behavior based detection is proposed using rule based genetic algorithm. Model is trained using UNSW NB15 dataset. This allow model to detect all types of modern attacks. Model overall performance is observed with 98.5% accuracy and 0.11% for false alarm. Keywords Hybrid intrusion detection system · Ensemble · Random forest · Genetic algorithm · Naïve bayes · Support vector machine

1 Introduction In era of virtual world an exponential growth of internet is marked. This is possible due to the reliability and availability provided by various internet providers now days. From last some decades a technical shift from traditional networks to multimedia networks is observed. Manan et al. [1] specifies the shift form text transmission to multimedia transmission. All the networks should be capable of transmission text, audio, video or any type of multimedia data. Due to increase in transmission of data security of data becomes utmost important issue in network security. Networks are transmitting data from one location to another, while this transmission data has to S. R. Khonde (B) Department of Computer Science and Engineering, Modern Education Society’s College of Engineering, SPPU, Pune, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_56

769

770

S. R. Khonde

pass through various networks. All networks have their security system to protect data. These systems can be firewalls or intrusion detection systems (IDS). Every organization has its own security policies that are applicable to data transmitting through that network. But sometimes if network is not secure or vulnerable data transmitting through it can be misled or modified. To upgrade towards new data transmission era there is a need of system to provide security to data. This is the main reason why IDS is an important part of any network now days. An IDS is a system which monitors both inside and outside packets. IDS analyze all the communication inside the network to avoid all malicious activities. It also analyzes each packet entering into the system. IDS generate alarm and store the information about that malicious packet. This alarm is send to administrator so that further action can be taken to secure network. IDS are mostly use for attack detection in networks. Nadiammai and Hemalatha [2] explain various data mining algorithms for attack detection. Almseidin et al. [3] provides evaluation of various algorithms of machine learning used for detection. Machine learning algorithms improve accuracy and detection rate of IDS. It helps to make IDS intelligent such that actions after attack detection can be taken by IDS. Deep learning can be considered as an emerging area in attack detection using IDS as proposed by Vinayakumar et al. [4]. IDS can perform attack detection using two important methods as explained by Butun [5]. Slight deviation in the usual behavior outline of the traffic can be deliberated as an anomaly attack. Alazab et al. [6] states that this type of IDS used to detect any type of novel attack even the web application or web services attacks.

2 Literature Survey For identifying intrusions in several networks intrusion detection system is used. Most of the IDS make use of standard datasets and various algorithms from machine learning to complete their task. Some of the IDS also use data mining techniques to complete work of detection. Most of the researcher provides approaches to be used like distributed or collaborative IDS. As various types of attacks are known researchers use different algorithms to improve performance such that detection of all types of attacks can be done. To increase enactment of anomaly based IDS Aburomman and Reaz [7] elaborated various ML algorithms. Hybrid and ensemble approaches are used by IDS to improve detection performance. Many ML and DM algorithms used for IDS with comparison are elaborated. Discussion and review of algorithms in machine learning is elaborated by Buczak and Guven [8]. They further elaborate parameters which can be used like time complexity, algorithm complexity and accuracy. This will help to select efficient technique for IDS implementation. Feature selection role and methods are explained by Qassim et al. [9] such that most suitable subset can be found to increase efficiency of the IDS. Packet header based method is introduced to analyze and classify traffic. This method selects features

Anomaly Based Intrusion Detection System Using Rule Based Genetic …

771

subset using feature selection technique for monitoring and analyzing anomaly intrusions. Vimala et al. [10] elaborates various ML algorithms. Ahmed et al. [11] provides survey of various anomaly detection techniques like statistical, clustering, classification and information theory. Feature reduction is done using combination of principal components analysis (PCA) and information gain (IG). This combination provides better result as compared to individual technique. This features are used with ensemble of SVM classifier, instance based learning algorithms and multilayer perceptron. Performance is tested on Kyoto2006 + along with other standard dataset as ISCX 2012 and NSL-KDD. Feng et al. [12] proposed a novel framework by combining SVM with ant-colony network. Kdd99 dataset is used for implementation in two parts as testing and training. Comparison of ensemble approach with individual classifier is elaborated. Results shows that ensemble approach gives better detection rate with reduce execution time. Li et al. [13] discussed a new hybrid approach to avoid bias detection of intrusion. Classifiers used are KNN clustering algorithm with binary classifiers. Binary classifiers are used to classify all known attacks. Remaining novel attacks are consider as outlier for cluster and handled by KNN clustering algorithm. Experiment performed on KDD dataset. Liu, et al. provides insight on ensemble approach [14]. Various approaches can be used for attack detection like deep learning [15–17], ensemble and cloud [18]. Neural network approach for detection of intrusions in network is mentioned by Wu et al. [19]. Various neural network algorithms are used by researchers now days for IDS. Authors make use of Computation Neural Network (CNN) for attack detection. IDS show better efficiency and good performance in terms of detection rate. Similar approach is use by Xiao, et al. [20]. They propose a method to convert traffic into image format. This will help to reduce computational cost of the system. Feature reduction is done using combination of encoding and principal component analysis. KDD99 dataset is use to perform experiments which results in high detection rate with reduced computational cost. Li et al. [21] propose a novel auto-encoder IDS with random forest. Random forest machine learning algorithm provides good detection accuracy but fails when improper dataset is used. Proposed method use deep learning auto encoding technique to improve performance of RF and reduce training and testing time. This method use feature selection and feature grouping to reduce number of features require for training. Results shows less training time of proposed method as compare to traditional method. IDS do not use single classifier for intrusion detection instead it make use of ensemble techniques to do so. In [22, 23] author makes use of feature selection technique with ensemble of classifier to improve performance of classifier. Experiment in both papers show that ensemble provides better performance as compare to individual classifier. Mukherjee and Sharma [24] explain importance of feature selection. Feature selection is done using information gain, correlation analysis and gain ratio. These three are compared with new novel method feature vitality based reduction.

772

S. R. Khonde

From literature survey we can observe that various approaches and frameworks helps is performance improvement of intrusion detection. Most of the ensemble techniques are use with feature selection for reducing computation time. Detection accuracy of IDS is depends on dataset used by it. Various datasets and its features are available in market. Various datasets are reviewed by Ring et al. [25]; it makes use of various dataset properties to analyze it. Moustafa et al. [26] gives detailed knowledge of various datasets used in intrusion detection.

3 Methodology of Proposed System The proposed model consists of an intrusion detection system capable of detecting all types of modern attacks. Modern IDS provides efficient performance and good detection rate. As most of the IDS works on signature based approach they were not able to detect modern attacks. Anomaly based detection for detection of unknown attacks is implemented in model. Anomaly detection is based on behavior checking of packets. Behavior analysis is done using genetic algorithm so that normal and malicious behavior of the system can be detected easily. If behavior is normal it is consider as a normal traffic and passed into the network. Otherwise if the behavior is malicious then packet is rejected by considering it as new attack. To check realism of model multiple classifiers are trained and tested individually along with ensemble approach. As per observations Ensemble approach gives good accuracy as compared to individual classifiers.

3.1 Ensemble Classifier To test model for modern attack detection various techniques from machine learning are used. These techniques are basically used to provide better accuracy and detection rate. Machine leaning classifier is divided into multiple classes as unsupervised, semi-supervised and last is supervised. This classification is based on the leaning methods classifiers use to train themselves. Model makes use of supervised learning classifiers to test dataset on the real time traffic. Ensemble approach helps to gain accuracy and performance of detection. All the three classifiers are tested in ensemble approach. Research of many researchers proves that ensembling of classifiers shows good performance of system as compare to individual classifier. Ensembling is done with majority voting algorithm in model. In this algorithm votes from each classifier is taken and final decision is made accordingly. Votes are nothing but predictions generated by every classifier, once packets are analyzed for attack. Classifiers used are supervised learning classifier, which makes use of labeled data to train and test classifier. Classifiers used are SVM, naïve Bayes and RF.

Anomaly Based Intrusion Detection System Using Rule Based Genetic …

773

3.2 Anomaly Detection: Genetic Algorithm This algorithm does attack detection on the basis of behavior. System behavior is described using rules. Behavioral rules are demarcated using genetic algorithms. Genetic algorithm use behavior patterns to define rules in four step process. It will first generate initial population followed by chromosome designing. For every chromosome, fitness value is calculated to design genetic operator. Rule based dataset is formed using process of genetic algorithm. Output of this process is rule based dataset. Input used for this is standard dataset UNSW NB15 for intrusion detection. UNSW NB15 dataset is used to find the optimal features which are used to find final rules. Each feature set is used to calculate fitness value. The one having strong fitness value will be converted into rules to create dataset. Further classifiers are trained sing this dataset. Naïve Bayes, random forest and SVM classifiers are used. Genetic algorithm process to define rule based dataset is explained below.

3.2.1

Rule Based Dataset

Each packet entering in network will be observed and checked according to behavior rules used to train classifiers. Normal behavior of data is defined using rules. Any deviation of rules is observed as abnormal behavior. Behavior match with rule will be considered as normal behavior. To create a rule based database from the standard benchmark datasets genetic algorithm follows four steps as explained below. Process basically starts with random selection of population from chromosomes. Chromosome is a problem need to be solved. According to various features of the problem, chromosomes are represented using bits, characters or numbers. These are represented at different position as required. These positions are randomly changed within a range of values and mostly known as genes. During a stage of process a set of chromosomes used is called population. As number of populations can be generated by randomly picking up genes from chromosomes an evaluation function is used to calculate its fitness. Goodness of population is obtained using fitness value which shows the importance of that population. Crossover and Mutation helps in generating more specific rules. Genetic algorithm process is described in Fig. 1. Initial formulation of chromosome is followed by fitness function evaluation. More specific chromosome is obtained using crossover and mutation. Rules are formed out of this every chromosome to generate database. Abnormal activity can be detected using this dataset. Process explained in Fig. 1 is a general process followed by genetic algorithms. The size of population, folds used for crossover and the rate of mutation based on benchmark datasets. This process helps in describing specific rules to define normal behavior of data. These rules are applied and used for training all the classifiers. Ensemble approach is used to find final prediction. Final prediction of the anomaly

774

S. R. Khonde

Fig. 1 Rule based database generation process using genetic algorithm

based detection engine will be given by majority voting. Rules from the database is represented as shown below. If {condition} then {action} else {action}. All rules are stored in above format in rule base dataset. Example of rule is given below. If {connection consist of: source IP address 122.17.2.28; destination IP address: 138.22.106.45; destination port number: 15; connection time: 08 s} then {stop the connection} else {allow connection to continue}. Explanation of rule is, if any connection request is coming in network from source IP address 122.17.2.28 having destination IP address as 138.22.106.45 with destination port number 15 and connection time required is 8 s then reject that transmission by stopping that connection. All the blocked IP addresses can be used in formation of rules. If above condition does not match then packets are consider as normal packets. Figure 2 shows the architecture of anomaly based detection engine. For genetic algorithm crossover rate consider is 0.15 along with 0.35 as mutation rate. Fitness functions used is shown in Eq. 1 below which is used to calculate fitness value for each chromosome. Generations used are depending on types of attacks from dataset. So it can be 5, 7 or 10 in order to increase accuracy of detection. Reason behind number of generations is because number of datasets used for testing. F= where p: count of attacks correctly classified

q p − P Q

(1)

Anomaly Based Intrusion Detection System Using Rule Based Genetic …

775

Fig. 2 Anomaly based detection engine

P: Count of attacks q: normal packets correctly classified in count Q: total number of packets in population. Input passed is the standard dataset UNSW NB15 for rule generation. These dataset.csv files as passed as input to genetic algorithm to get rule based database. These rules are used to train classifiers in terms of behavior of data. Rules are used to define normal behavior of data. Any deviation will be observed as abnormal against the rule. This data is announced as attack. Upon detection of attack alert is generated for administrator. All packets observed as normal packets are transfer. An ensemble approach using majority voting is used to avoid biased prediction.

4 Results and Discussions For experimental setup UNSW NB15 is used to train classifier. Dataset allows detecting many modern attacks. Classifiers are tested in real time where packets are captured using Wireshark network analyzer. All the packets from network are passed to handling and detecting attacks using anomaly based approach. IDS performance is evaluated using parameters as FAR, Accuracy and DR. All three classifiers are used to test dataset into real time environment. Various parameters used for testing is explained below.

776

S. R. Khonde

Testing of all classifiers using UNSW NB15 dataset is done according to ensemble approach. Formula to calculate parameters accuracy, detection rate and false alarm rate is represented in Eqs. (2–4). accuracy =

TP + TN total

(2)

DR =

TP FN + TP

(3)

FPR =

FP TN + FP

(4)

where TP—Correct positive prediction. FP—Incorrect positive prediction TN—Correct negative prediction FN—Incorrect negative prediction. Model makes use of all parameters discuss above to evaluate performance of classifier. Classifier used is SVM, naïve bayes and RF. This all classifiers work as supervised learning classifier to do validation of dataset and evaluating performance of classifier. The classifier performances are compared according to all parameters and the best classifier is used for implementation of IDS. Performance of classifiers is check by training all with UNSW NB15 dataset to detect modern attacks. While testing for attacks confusion matrix is created such that parameters can be calculated according to dataset. Following section represents confusion matrix for dataset.

4.1 Performance Using UNSW NB15 This dataset is generated by cyber security research group at the Australian Centre for Cyber Security (ACCS) in 2015. This dataset consist of normal records with records of nine different type of attack. UNSW-NB15 is one of the most complex dataset as compare to other datasets. This dataset sets a benchmark in the area of NIDS for datasets. Total size of dataset is 100 GB. Pcap files of 100 MB are generated to provide reliable features in the dataset. Table 1 shows parameter evaluation over UNSW NB15 dataset. Graphical representation of results is shown in Fig. 3.

Anomaly Based Intrusion Detection System Using Rule Based Genetic …

777

Table 1 Performance of model on accuracy, detection rate and false alarm rate Attacks/parameters

Accuracy

Detection rate

False alarm rate

Normal

0.94

0.90

0.33

Generic

0.99

0.95

0.31

Exploits

0.89

0.80

0.38

Fuzzers

0.94

0.92

0.61 0.25

Reconnaissance

0.98

0.95

DoS

0.91

0.86

0.64

Analysis

0.96

0.92

0.91

Backdoor

0.97

0.91

0.60

Shellcode

0.92

0.85

0.27

Worms

0.98

0.95

0.20

1.2 1 0.8 0.6 0.4 0.2 0

Accuracy

Detection Rate

False Alarm Rate

Fig. 3 Performance parameters for all classes over UNSW NB15 dataset

5 Conclusion The presented model uses hybrid model of anomaly based detection technique and genetic algorithm for attack detection. Behavior based analysis is done using anomaly based detection based on genetic algorithm. Rules are generated using genetic algorithm during packet analysis. Prediction given by anomaly detection will be the final prediction. Model makes use of ensembling technique to increase accuracy of prediction and rate of detection along with reduce false rate. Model performance is checked using parameters as false positive rate, detection rate and accuracy. Testing of this model is done in real time environment. To detect modern era attacks model is tested using UNSW NB15 dataset. Overall performance of model is enhanced as compared to existing hybrid models. Limitation of this model is it will not detect attack hidden

778

S. R. Khonde

in the payload of protocol. In future work a new phase can be added to this model based on content analysis so that maximum type of attacks can be detected using this novel method. This model can also be provided as service to customers such that depend on network attack users can request any type of detection engine to used, signature, anomaly or content based detection engine.

References 1. MananJ, Ahmed A, Ullah I, Merghem-Boulahia L, Gaiti D (2019) Distributed intrusion detection scheme for next generation networks. J Netw Comput Appl 147 2. Nadiammai G, Hemalatha M (2014) Effective approach toward Intrusion detection system using data mining techniques. Egypt Inform J 15:37–50 3. Almseidin M, Alzudi M, Kovacs S, Alkasassbeh M (2017) Evaluation of machine learning algorithms for intrusion detection. In: 15th International symposium on intelligent systems and informatics, Subotica, Serbia, pp 14–16 4. Vinayakumar R, Alazab M, Soman K, Poornachandran P, Al-Nemrat A, Venkatraman S (2019) Deep learning approach for intelligent intrusion detection system. IEEE Access 7:14525–41550 5. Butun I, Morgera S, Sankar R (2014) A survey of intrusion detection systems in wireless sensor networks. IEEE Commun Surv Tutorials 16(1):266–282 6. Alazab A, Hobbs M, Abawajy J, Khraisat A, Alazab M (2014) Using response action with intelligent intrusion detection and prevention system against web application malware. Inf Manage Comput Secur, 22(5):431–449 7. Aburomman, Reaz M,”A survey of intrusion detection systems based on ensemble and hybrid classifiers. Comput Secur 65:135–152 8. Buczak, Guven E (2016) A survey of data mining and machine learning methods for cyber security intrusion detection. IEEE Commun Surv Tutorials 18(2):1153–1176 9. Qassim Q, Zin A, Aziz M (2016) Anomalies classification approach for network-based intrusion detection system. Int J Netw Secur 18(6):1159–1172 10. Vimala S, Khanaa V, Nalini C (2019) A study on supervised machine learning algorithm to improvise intrusion detection systems for mobile ad hoc networks. Clust Comput 22:4065–4074 11. Ahmed M, Mahmood AN, Hu J (2016) A survey of network anomaly detection techniques. J Netw Comput Appl 60:19–31 12. Feng W, Zhang Q, Hu G, Huang JX (2014) Mining network data for intrusion detection through combining svms with ant colony networks. Futur Gener Comput Syst 37:127–140 13. Li L, Yu Y, Bai S, Hou Y, Chen X (2017) An effective two-step intrusion detection approach based on binary classification and k-NN. IEEE Access 6:12060–12073 14. Liu J, He J, Zhang W, Ma T, Tang Z, Niyoyita JP, Gui W (2019) ANID-SEoKELM: adaptive network intrusion detection based on selective ensemble of kernel ELMs with random features. Knowl Based Syst 177:104–116 15. Khonde SR, Ulagamuthalvi V (2022) Blockchain: secured solution for signature transfer in distributed intrusion detection system. Comput Syst Sci Eng 40(1):37–51 16. Khonde SR, Ulagamuthalvi V (2022) Hybrid intrusion detection system using blockchain framework. Eurasip J Wirel Commun Netw 58 17. Ferrag MA, Maglaras L, Moschoyiannis S, Janicke H (2020) Deep learning for cyber security intrusion detection: approaches, datasets, and comparative study. J Inf Secur Appl 50:102–419 18. Garg S, Kaur K, Batra S, Aujla GS, Morgan G, Kumar N, Zomaya AY, Ranjan R En-abc: an ensemble artificial bee colony based anomaly detection scheme for cloud environment. J Parallel Distrib Comput 135:219–233 19. Wu K, Chen Z, Li W (2018) A novel intrusion detection model for a massive network using convolutional neural networks. IEEE Access 6:50850–50859

Anomaly Based Intrusion Detection System Using Rule Based Genetic …

779

20. Xiao Y, Xing C, Zhang T, Zhao Z (2019) An intrusion detection model based on feature reduction and convolutional neural networks. IEEE Access 7:42210–42219 21. Li X, Chen W, Zhang Q, Wu L (2020) Building auto-encoder intrusion detection system based on random forest feature selection. Comput Secur 95 22. Zhou Y, Cheng G, Jiang S, Dai M (2020) Building an efficient intrusion detection system based on feature selection and ensemble classifier. Comput Netw 174 23. Rajadurai H, Gandhi UD (2020) A stacked ensemble learning model for intrusion detection in wireless network. Neural Comput Appl 24. Mukherjee S, Sharma N (2012) Intrusion detection using naïve bayes classifier with feature reduction. Procedia Technol 4:119–128 25. Ring M, Wunderlich S, Scheuring D, Landes D, Hotho A (2019) A survey of network-based intrusion detection data sets. Crypt Secur 26. Moustafa N, Saly J (2015) UNSW-NB15: a comprehensive data set for network intrusion detection systems (UNSW-NB15 network dataset). In: Military communications and information systems conference, pp 10–12

Hybrid Learning Approach for E-mail Spam Detection and Classification Rimitha Shajahan and P. L. Lekshmy

Abstract Email is the most common channel through which cyber attackers commit crimes and initiate spamming attacks. Spamming is a popular method of sending unsolicited messages in order to distribute malware. The disadvantages of spamming include system slowdown, time consumption, and the presence of viruses. The main goal of this work is to identify spamming text, url, and message id and reduce the rate of spamming on e-mail. To reduce and control spam in e-mail, the text content, links, and header information in the corresponding e-mail can be analyzed. In the above scenarios, one of the main driving forces is the use of natural language processing to distinguish between spam and non-spam occurrences. The major goal of the proposed study is to consider three features—text, url, and message id—rather than deploying a single feature. The deep learning models, which play a significant role in predicting the spamming texts than any other traditional machine learning models. The machine learning models also play an important role in predicting the spamming urls and message id. Three standard datasets—Enron for text content, Phishtank for urls and SpamAssassin for Message Id are collected and processed. The aggregation of machine learning and deep learning models are an added advantage in this work. This identifies the spamming e-mail on the real-world datasets by using LSTM, Random Forest, Multinomial Naive Bayes models, which then evaluates the performance. Further, the output from the three models namely text, url, message id are integrated using weighted fusion approach and finally the output will be obtained as spam or ham. Keywords Long short term memory · Random Forest · Term frequency-inverse document frequency · Bag of words

R. Shajahan (B) · P. L. Lekshmy LBS Institute of Technology for Women, Poojappura, Thiruvananthapuram, Kerala, India e-mail: [email protected] P. L. Lekshmy e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_57

781

782

R. Shajahan and P. L. Lekshmy

1 Introduction E-mail is a type of electronic communication in which data is stored on a computer and exchanged by establishing a telecommunication connection between two users. Email is a message, which include text, files, photos, or other attachments and is sent through a network to a single recipient or a group of recipients. Spamming is the way of transferring a large number of people several unwanted messages for the goal of commercial advertising, non-commercial proselytizing, any forbidden purpose, or simply delivering the same message to a huge number of people. Spam attacks are usually communicated via electronic communications includes texts and emails. Spamming has indeed demonstrated to be an effective attack, misleading a diverse group of people. Attackers frequently pose as popular social media sites, banks, IT managers, or popular commerce websites. These emails may persuade recipients to click on links that lead to malware downloads or to enter personal information on a malicious website that appears to be authentic. Spamming is a deceitful technique that uses social engineering and technology to get private information, including passwords and credit card details, by pretending to be a reliable individual or business in an electronic conversation. Spamming is the practise of sending fabricated emails that look real and come from reliable sources, including financial organisations or e-commerce websites, in order to convince recipients to visit shady websites by clicking on links included in the spam email. Email spam has become a serious issue in recent years, and as internet users grow, so does the spam email population. They are being misused for dishonest and unethical activities like fraud and spamming. Sending spam emails with harmful links in them, which can compromise our system and access yours. Spammers can rapidly set up a bogus profile and an email account, and they can pose as real persons in their spam emails. The targets of spam are people who are not aware of the schemes. In addition to the annoyance and time lost sifting through unwanted emails, spam can seriously affect users’ computers by infecting them with malicious software that can impair the system and steal personal information. It has the power to use up all available network resources. Millions of internet users are impacted by spam attacks, which are extremely expensive for businesses and spam recipients. Spam has grown to be a serious hazard to both individuals and businesses. In order to ensure the greatest possible level of human benefit and security, strive to detect spamming emails based on the text content, URL present in the email, and message-Id of the email.

2 Related Works The review of literature focus on E-mail spam detection and classification using Machine Learning and Deep Learning techniques based on the three features—text content, url and message id of the e-mail.

Hybrid Learning Approach for E-mail Spam Detection …

783

2.1 Machine Learning Techniques Joshi et al. [1] proposed an ensemble machine learning in order to detect fraudulent URLs communicated in emails. Blacklists are a popular and established approach of identifying dangerous URLs. The classification of URLs has been studied using machine learning and deep learning models that integrate static data such as lexical characteristics from the URL string, host information. It shows how to categorise URLs using only static lexical information from the URL string. In this investigation, the Crisp-DM Methodology was employed for analyse the presence of spam and ham urls. The classification using Random Forest model’s findings with 92% accuracy outperforms than Naive Bayes, SVM, and Logistic Regression, ensemble classifiers like AdaBoost and Gradient Boost. Sultana et al. [2] investigated that e-mail has been the most significant mode of communication in recent years and spam is a serious problem that is not only inconvenient for recipients. Kaggle’s SMS spam was used as the dataset. The methodology presented in this work not only detects the spam term but also the system’s IP address, enabling the spam message to be immediately recognised as banned the next time it is transmitted from that system based on the IP address. Siddique et al. [3] focused on as the social communication has progressed that an e-mail has remained the most usual methods of formal and informal communication. Urdu, a South Asian language spoken largely in Pakistan, is gaining traction as a communication medium on social media platforms, websites, and emails. As the emails usage has grown, so has the volume and various types of spam messages in Urdu is observed. The python script Google-trans, which leverages the Google Translate Ajax API, was used to collect 5000 emails from the web resource ’kaggle’ and translate them into Urdu. This work proposes leveraging existing machine learning techniques such as Naive Bayes, CNN, SVM, and LSTM to recognise and categorise e-mail content. LSTM model surpasses other models with accuracy score of 98.4%. Patgiri et al. [4] proposed to do a detailed analysis into the detection of malicious URLs using Machine Learning methods. To extract these lexical properties, the URL is broken down into a series of words or tokens depending on the delimiters in the URL. The length of the URL is obtained by reading the URL as a string and instantly calculating the length of the string as an input to the Feature Extraction(URL) method. Host-based features are the URL’s host-name properties are used to acquire host-based functionality. Since identifying malicious URLs is a binary classification problem, several machine learning algorithms, including Random Forests, SVMs, and Naive Bayes, are used to the training dataset. Furthermore, it has been found that for the given problem, the Random Forest classifier performs better than the SVM classifier. Washha et al. [5] present as long as spammers continue to develop new ways and tactics for defeating and confusing email spam filtering systems. The header session messages, which are based on publicly available datasets, are utilised to demonstrate a useful and powerful email header feature. To confirm the effectiveness of the acquired

784

R. Shajahan and P. L. Lekshmy

header features in filtering spam and ham broadcasts, the impact of various machine learning-based classifiers on the retrieved header characteristics is again assessed and compared. Using the following classifiers: Random Forest (RF), C4.5 Decision Tree (J48), Voting Feature Intervals (VFI), Random Tree (RT), REPTree (REPT), Bayesian Network (BN), and Naive Bayes. Random forest outperforms than other classifiers and filter out the spam and ham header information.

2.2 Deep Learning Techniques AbdulNabi and Yaseen [6] aims the work is to put state-of-the-art models to the test with the goal of detecting spam emails. Spambase data set from the UCI machine learning repository and Kaggle’s open source Spam filter dataset are used. The best model, with an accuracy of 98.67% and an F1 score of 98.66%, is the bert-base-cased transformer model because it uses attention layers to take context into consideration. When compared to Keras word embedding, which assigns a different number to each word, Bert contextual word embedding enhances the capacity to identify spam emails. It was used in the BiLSTM model, which has an F1 score of 96% and accuracy of 96.43%. The results versus unknown data show the robustness and endurance of the fully fitted models. Jain et al. [7] investigated that spam classification is a prominent topic in natural language processing, especially using the internet for social networking grows as the number of people increases. LSTM can learn abstract traits, unlike conventional classifiers that need manually created features. Before being input into the LSTM for classification, the text is transformed into semantic word vectors using word2vec, WordNet, and ConceptNet. The classification’s outcomes are contrasted against benchmark classifiers including SVM, Naive Bayes, Artificial Neural Network, k-nearest Neighbor, and Random Forest. Two datasets—the SMS Spam Collection dataset and the Twitter dataset—are used to compare the outcomes. Results are assessed using the accuracy and F-measure. The outcomes show that LSTM can significantly outperform current machine learning methods for spam identification. Bhuvaneshwari et al. [8] suggested that consumer reviews are a valuable source of information in the world of e-commerce. In this study, a deep learning (DL)based novel framework for learning document level representation for recognising spam reviews as a replacement for ML-based detection is described. Using the Self Attention-based CNN Bi-LSTM (ACB) mechanism, the approach assesses the significance of each word in the phrase and scans the text for signs of spamming. The model then use a convolution neural network to learn phrase representation and extract higher-level n-gram features (CNN). In order to identify spam reviews with context, sentence vectors are concatenated as document feature vectors using Bi-directional LSTM (Bi-LSTM). Fariska [9] proposed the study of the effect of using various combinations of pre-processing steps on some spam detection algorithms. They selected two spam detection classifiers that represented different approaches: Naïve Bayes and Support

Hybrid Learning Approach for E-mail Spam Detection …

785

Vector Machine. Pre-processing methods are differentiated in this study are: noise removal, stemming, lemmatization, and term frequency (TF-IDF). Each classifier’s accuracy will increase if the proper pre-processing techniques are applied (or not applied) to it. For the Naive Bayes classifier, the stop words removal and stemming combination performs better than other combinations. The pre-processing stage, on the other hand, does not always result in improved classification results for Support Vector Machine (SVM) classifiers. Word shapes and the existence of stop words are sensitive to the probabilistic classifier—Naive Bayes classifier. SVM, on the other hand, does not require nearly all pre-processing methods because it is a nonprobabilistic classifier.

3 Proposed Work Detecting spam email is a difficult task because it is skillfully made to appear as real. Aside from attachments, an email primarily consists of a header, text body and links. In the proposed system, three features are used to detect and classify whether it is a spamming or a legitimate email. The proposed system makes use of Aggregated Model for E-mail Spam Detection and Classification. In this work, Enron dataset for text content, Phishtank dataset for URLs and SpamAssassin dataset for message id are utilized. First of all, read the data and then it will undergo the data pre-processing that contains removing noisy data, replacing missing data and removal of stopwords, punctuations etc. After that, feature extraction method is used for generating the relevant features and the input is given to the aggregated model. Thus detect and classify the email as spam or ham (Fig. 1).

3.1 Detecting Spam E-mail Based on Text Content The proposed model for E-mail Spam Text Content Analysis contains the steps as follows data collection, pre-processing, feature extraction and classification and finally the spam text content indicated as 1 and ham text content indicated as 0: Text Content Dataset: The data is collected from Enron Corpus dataset that is available

in Kaggle. The dataset contains 5728 text contents that is about 4360 ham and 1360 spam text contents. About 500,000 emails generated by Enron Corporation employees were included in Enron email dataset. Real email is included in the dataset, which is openly accessible and can be used to improve the present email tools. Dataset Preprocessing: After collecting the dataset, the preprocessing is performed. The

available data is converted into lower case and punctuations are removed. The dataset includes the symbols (hash tags), @ (user mention), and rt (re-tweets), as well as the URLs http and https. After removing these symbols, word tokenization is carried out. Sentences are broken down into tokens through the process of tokenization.

786

R. Shajahan and P. L. Lekshmy

Fig. 1 Hybrid learning approach for spam e-mail detection and classification

Here, a considerable sample of text is divided into words using word tokenization. Stop words are finally eliminated and these are phrases that have no influence on the detection of email spam. Stop words removal is the procedure of getting rid of these stop words. The Natural Language Tool Kit has 179 stop words in English that is used in this work. Feature Extraction using Bag of Words Model: The bag-of-words model is used to convert

the text into numerical form that is available in natural language processing and information retrieval. Depict a sentence as a vector of words, much like the term itself. The most frequent words are obtained from the text and use a dictionary to hold the bag of words. Tokenize the text to words and then every word in text check whether the word exists in declared dictionary. If the word is present, increase the count by one and else hold the word in dictionary and set the count as 1. For building the bag of words, make a vector that will check a word in every text is a frequent word or not. If the word is a frequent word, mark it as 1 otherwise mark it as 0. Classification using LSTM Model: After the feature extraction step, the input is fed

into the deep learning model architecture for e-mail spam analysis on Enron dataset. The dataset is split into training for 70% and testing for 30%. LSTMs perform considerably better since they are skilled at remembering particular patterns. The important information is saved and all the irrelevant information is deleted in every single cell when the LSTM goes through each hidden layer, just like every other NN. The model consists of 2 simple rnn layer, two dropout layers, LSTM layer, 2 dense layer and elu layer and dropout layer. The Input layer takes sequence of inputs. Each

Hybrid Learning Approach for E-mail Spam Detection …

787

word in the input post is transformed into a word embedding before being tokenized. The following layer receives these embedding vectors as input. All of the words in the training dataset will have an embedding learned by the Embedding layer, which starts out with random weights. The words are transformed into word embeddings, which are then transformed into lower-dimensional vectors [10]. Word embeddings are best used for deep learning and NLP applications since they are quick and effective and can capture context similarity due to their smaller dimensionality. Each word in the phrase has its word embeddings concatenated and supplied to the hidden layer. A predetermined size sequence of words is processed by the embedding layer. One drop out layer is given after the LSTM layer to reduce the overfitting. The dropout rate for the dropout layer is set as 0.2. Here the LSTM has 25 layers. From the input vectors, features are extracted. LSTM extracts features from embedding vectors. The network’s dense layer 1 is the top layer, and it flattens and integrates the high-level information that the other layers have learned. The regularisation dropout layer is also a part of this layer. The Dense layer2 depends on number of classes here 2, spam and ham. Finally, the Activation function SoftMax is applied to take probability. The SoftMax output layer receives this layer’s output for prediction that is spam text content indicated as 1 and ham text content indicated as 0. Adam optimizer is one of the most efficient optimization algorithms used in the proposed model. It optimizes model performance mostly by reducing the cost function associated with the model. Adam optimizer is easy to implement, requires little memory space, performs well with huge data sets and large parameters, as well as situations with noisy or sparse gradients. The lstm model is to choose from a wide range of LSTM parameters, including learning rates and input and output biases. Thus, there is no need for precise modifications. With LSTMs, updating each weight is simpler than with Back Propagation Through Time (BPTT), reducing complexity to O(1). Tried out with Bi-lstm model for the classification, it gives less accuracy and false positive rate is high and thus selected the lstm model for classification with high accuracy, less loss rate and false positive rate.

3.2 Detecting Spam E-mail Based on URL The proposed model for E-mail Spam URL Analysis contains the steps as follows data collection, pre-processing, feature extraction and classification and finally the spam url indicated as 1 and ham url indicated as 0: URL Dataset: Url dataset is collected from phishtank dataset and other datasets from

github. The labeled dataset contains 10,000 urls that is 5000 for spam urls and 5000 for ham urls. Data Preprocessing: In the data preprocessing step, remove the reduntant data and

eliminate the missing data.

788

R. Shajahan and P. L. Lekshmy

URL features: URL is splitted into protocol and domain address part. To extract the

url features, the URL is broken down into a series of words or tokens depending on the delimiters in the URL and check whether the url having the below mentioned features: 1. Presence of IP address: Users can be sure that someone is attempting to steal their personal information if an IP address is used in the URL instead of the domain name, such as “http://125.98.3.123/fake.html.” 2. URL length: Spammers can hide the suspicious portion of a URL in the address bar by using a lengthy URL. Shortening Service: On the “World Wide Web,” a URL can be significantly shortened while still directing to the desired webpage by using this technique. This is done by using a “HTTP Redirect” on a short domain name that links to the full URL of the webpage. The URL “http://portal.hud.ac.uk/” can be condensed to “bit.ly/19DXSk4”, for instance. 3. Presence of @ symbol: Using the @ sign in a URL causes the browser to ignore everything before the @ symbol, and the actual address frequently comes after the @ symbol. 4. Double slash redirecting: The visitor will be moved to another website if the URL path contains the character “//.” 5. Prefix-suffix: Legitimate URLs rarely employ the dash symbol. In order to give users the impression that they are visiting a trustworthy website, spammers frequently append prefixes or suffixes to the domain name, separated by (–). Feature Extraction using TF-IDF Model: TF-IDF method is used to identify the important

term in the corpus and focus on that word for detecting the presence of spam and ham. This method makes tokens after splitting by double slash redirecting (//), dash (–), dot (.), remove .com and reduntant tokens. A numerical measure called term frequency-inverse document frequency aims to show how significant a term is to a given document in a collection or corpus. A term’s weight in a document is merely proportionate to how often it appears. t f (t, d) = count o f t in dnumber o f wor ds in d

(1)

Document Frequency, or df, is the quantity of times the term t appears in the document collection N. d f (t) = occurr ence o f t in documents

(2)

Inverse Document Frequency (IDF) is the dataset’s size N divided by frequency of the text df(t). id f (t) = log(N /d f (t)) (3) tf-idf(t, d) indicates how important the term t in the whole document d, tf(t, d) indicates the number of times the term t occurs in the document d and idf(t) indicates the size of the corpus to the frequency of the text t.

Hybrid Learning Approach for E-mail Spam Detection …

t f − id f (t, d) = t f (t, d) ∗ id f (t)

789

(4)

The TF-IDF value in this scenario is to range above 0.5 which considers the word that is more needed and below that range can be discarded. After the feature extraction using TF-IDF, get the feature vector and that will be passed as input for the classification using Random Forest. In order to account for the fact that some words are used more frequently than others, the relevance of the TF-IDF value rises according to the number of times a term appears in the text and is offset by the number of documents in the corpus that contain the word. Classification using Random Forest: The dataset is divided into training for 70% and

testing for 30%. It also marks the presence of each feature in the Url using 1 or 0 and TF-IDF value that will be given as input to the random forest model. Random forest is the model in which features having its own decision trees and using majority voting mechanism final output predicts as spam (1) or ham (0) [11, 12].

3.3 Detecting Spam E-mail Based on Message Id The proposed model for E-mail Spam Message Id Analysis contains the steps as follows data collection, pre-processing, feature extraction and classification and finally the spam message id indicated as 1 and ham message id indicated as 0: Message Id Dataset: SpamAssassin dataset is collected from Kaggle. The dataset con-

tains 4008 message ids that is extracted from easy ham, hard ham and spam text files and contains 1399 spam and 2609 ham message ids. Spam message id is denoted using 1 and non-spam message id is denoted using 0. Preprocessing of Message Id: Spam assasin dataset has easy ham, hard ham and spam

dataset of message ID. From this easy ham and hard ham message ID are taken. LHS and RHS were separated using @ and data is labelled using 1 or 0. Two folders were created, one containing spamming message id and other containing non spamming message id (easy ham and hard ham). Feature Extraction using N-gram Frequency: N-gram frequency (initially bi-gram) is

done on the data using n-gram library from nltk package. It is a string or text that contains n characters in a row. For instance, in the word abc123, the character sequences for the 1st gram would be a, c, 2, etc. Similar to 2-grams, 2-grams are sequences of 2 characters that overlap, such as ab, bc, c1, 12, and 23. In a similar way, this concept can also be used to higher order n-grams [13]. Classification using Multinomial Naive Bayes: The dataset is divided into training for

70% and testing for 30%. Email filtering with naive Bayes classifiers is a common statistical technique. To determine if an email is spam or not, the Bayes theorem is applied after correlating token usage. The filter will modify the likelihoods that each

790

R. Shajahan and P. L. Lekshmy

word will appear in spam or valid email in its database for all terms in each training email. These word probabilities are used to determine whether an email that contains a specific set of words falls within one of the two categories—spam or ham message ids. P(W/S) is the frequency of messages containing a specific term in the messages labelled as spam message id. P(W |S) = P(W ) ∗ P(S|W )/P(S)

(5)

P(W/H) is the frequency of messages containing a specific term in the messages labelled as ham message id. P(W |H ) = P(W ) ∗ P(H |W )/P(H )

(6)

Each word in the email affects its likelihood of being spam, or merely the most intriguing words. The Bayes theorem is used to calculate this contribution, which is known as the posterior probability. The spam likelihood of the email is then calculated over all of its terms, and if the sum is higher than a predetermined threshold, the filter will classify the email as spam otherwise ham.

3.4 Model Aggregation For the aggregated model, the three features—text content, url and message id corresponding datasets are combined into single dataset of about 4008 data and each feature uses its own data pre-processing and feature extraction steps. The dataset is split into training and testing datasets. The training is about 70% and testing about 30% of the datasets. Further it will detect and classify the email as spam or ham using machine learning and deep learning techniques. The text content model is utilized the Enron dataset. When the e-mail text content is passed as an input to the text model. First, it will perform data pre-processing and then extracted the features using Bag of Words model. Finally, detection and classification is done using LSTM model and predicts the output as spam or ham text content. If text probability and predict is 1, the output predicted as spam text content else output as ham if text content text probability and predict is 0. The URL model is utilized the Phishtank dataset. When the e-mail attached URL is passed as an input to the URL model. First, it will perform data pre-processing and then extracted the features using TF-IDF model. Finally, detection and classification is done using Random Forest model and predicts the output as spam or ham URLs. If URL probability and predict is 1, the output predicted as spam URL else output as ham URL if URL probability and predict is 0. The Message Id is utilized the SpamAssassin dataset. When the e-mail header information—Message Id is passed as an input to the Message Id model. First, it will perform data pre-processing and then extracted the features using N-grams model.

Hybrid Learning Approach for E-mail Spam Detection …

791

Finally, detection and classification is done using Multinomial Naive Bayes model and predicts the output as spam or ham Message Ids. If Message Id probability and predict is 1, the output predicted as spam Message Id else output as ham Message Id if Message Id probability and predict is 0. The three models gives an output that will be either spam (1) or ham (0). Finally the aggregated model is done on the corresponding result of the three features using weighted fusion model and thus final output is obtained. Weighted fusion model is firstly find the prediction probability and the weight of each feature. Then the probability of the aggregated model is calculated as pb =



wi ∗ pbi

(7)

pb is the probability distribution of the model and wi is the average weight of the model and pbi is the probability distribution individual model. The probability distribution of the aggregated model will be generated and if the prediction is 1 (e-mail is spam) else the prediction is 0 (e-mail is ham).

4 Result Analysis For the aggregated model, the three datasets corresponding to text content, url and message id into single dataset and each feature has its own data preprocessing and feature extraction steps. Further it will detect and classify the email as spam or ham using machine learning and deep learning techniques. The three models gives an output that will be either spam (1) or ham (0). Finally the aggregated model is done using weighted fusion model and final output is obtained. The sample output for spam aggregated model is shown in Fig. 2. The sample output for ham aggregated model is shown in Fig. 3.

Fig. 2 Sample output of aggregated model indicates spam

792

R. Shajahan and P. L. Lekshmy

Fig. 3 Sample output of aggregated model indicates ham

Fig. 4 Confusion matrix for aggregated model

A confusion matrix is used to evaluate the performance of a classification algorithm. The true positive is 1253 that means the output is correctly predicted ham. Here the true negative is 418 that means the output is correctly predicted spam. A false negative is an outcome where the model falsely predicts the ham class. Here the false negative is 47 that means the output is falsely predicted spam (Fig. 4). A Classification report is used to evaluate the rate of predictions from a classification algorithm. After classification, a classification report was generated to analyze the precision, recall, f1-score and accuracy of the model (Fig. 5). The performance measures like precision, recall, f1-score and accuracy is evaluated for each individual models and aggregated model. The aggregated gives the better result with less false positive rate and correct prediction. Accuracy of the aggregated model is calculated as the total number of actual positives to the overall predictions. Precision of the aggregated model is calculated as the total number of actual positives to the total number of positives and number of negatives. Hybrid learning approach gives better accuracy as compared to the individual models for the prediction and evaluate the spam e-mail and ham e-mail (Fig. 6).

Hybrid Learning Approach for E-mail Spam Detection …

793

Fig. 5 Classification report for aggregated model

Fig. 6 Comparison of individual models with aggregated model

5 Conclusion Millions of internet users are affected by spam attacks, which are extremely expensive for businesses and spam recipients. Spam has recently been a major source of worry on social, economic, political and organizational levels, as it diminishes employee productivity and increases network traffic congestion. The various spamming operational modes serve as a reminder to focus on certain elements that could effectively identify spamming attempts. In order to solve the issue of spamming via email, a framework for successfully detecting spam has been suggested. It makes use of features that have been shown to be effective in the literature and produces high accuracy utilizing machine learning and deep learning approaches. By alerting

794

R. Shajahan and P. L. Lekshmy

users about emails that are associated with spamming, this framework assists users to avoid being spammed via emails. The aim to resolve this problem by proposing a system for the detection of spamming emails by considering three features for classifying an email as spam or non spam. The system focus on message id, message content and attached URL for detecting a spamming email using machine learning and deep learning techniques. The model is built for each of the three features and then integrated it for the final classification. Since the spammers are so skilled that they can easily fool the users by sending spam emails in a real way, the aggregated model helps the user to know whether the email is spam or not. Here False positive rate is reduced, accuracy is increased and chances of mis-classification of email is very rare.

References 1. Joshi A, Lloyd L, Westin P, Seethapathy S (2019) Using lexical features for malicious URL detection—a machine learning approach. ArXiv http://orcid.org/abs/1910.06277 2. Sultana T, Sapnaz KA, Sana F, Najath J (2020) E-mail based spam detection. Int J Eng Res Technol (IJERT) 9(06) 3. Siddique ZB, Khan MA, Din IU, Almogren A, Mohiuddin I, Nazir S (2021) Machine learningbased detection of spam e-mails. Article ID 6508784. Hindawi. https://doi.org/10.1155/2021/ 6508784 4. Patgiri R, Katari H, Kumar R, Sharma D (2019) Empirical study on malicious URL detection using machine learning. In: International conference on distributed computing and internet technology ICDCIT 2019: distributed computing and internet technology, pp 380–388 5. Washaha M, Khater IM, Qaroush A (2012) Identifying spam e-mail based-on statistical header features and sender behavior. In: International information technology conference and exhibition (CUBE), September 2012, Pune, India 6. AbdulNabi I, Yaseen Q (2021) Spam detection using deep learning techniques. In: The 2nd international workshop on data-driven security (DDSW 2021), March 23–26, Warsaw, Poland 7. Jain G, Sharma M, Agarwal B (2018) Optimizing semantic LSTM for spam detection. Int J Inf Technol 11(3) 8. Bhuvaneshwari P, Rao AN, Robinson YH (2021) Spam review detection using self attention based CNN and bi-directional LSTM. Multimed Tools Appl 80:18107–18124 . https://doi.org/ 10.1007/s11042-021-10602-y 9. Fariska R (2019) Study on the effect of preprocessing methods for spam email detection. Indonesian J Comput (Indo-JC) 10. Rahman SE, Ullah S (2020) Email spam detection using bidirectional long short-term memory with convolutional neural network. In: 2020 IEEE region 10 symposium (TENSYMP). IEEE, pp 1307–1311 11. Govil N, Agarwal K, Bansal A, Varshney A (2020) A machine learning based spam detection mechanism. In: 2020 Fourth international conference on computing methodologies and communication (ICCMC). IEEE, pp 954–957 12. Chen J-Y, Wang Y-J (2022) Semi-supervised fake reviews detection based on AspamGAN. J Artif Intell 4(1):17–36 13. Balasubramaniam V (2021) Design of associate content based classifier for malicious URL prediction by rule generation algorithm. J Inf Technol Digit World 3(1):44–56

Smart Solid Waste Management System Using IoT Technology: Comparative Analysis, Gaps, and Challenges Meenakshi Shruti Pal and Munish Bhatia

Abstract With the consistent spike in the world’s population and continued expansion of urban cities, waste is a highly visible city problem for every prominent stakeholder including the respective State Government of the territory. Waste management in the cities has big expenditures based on the method of waste collection and disposable systems that are opted by a particular city. There are several ways to enhance the waste collection mechanism, however, one of the most appropriate systems with appropriate use of technology is key to getting the waste management system right and making it commercially viable. The latest entrant to smart city waste management is the Internet of Things (IoT). The smart bins comprising IoT sensors and route optimization improve traditional waste management processes in the most efficient manner. Therefore, this paper reviews various IoT-based solutions used for waste collection and route optimization methods for garbage collector vehicles in the past decade. The objective of the paper is to study and analyze the state-of-the-art techniques used for the waste management system with smart IoT-enabled bins. In this paper, analyses of research papers are described in the literature. Research gaps from an existing work have been concluded based on the results of the study. Further, this paper also describes the various challenges and issues of the smart waste management system. This thus, calls for further improvement and innovation. Keywords Waste management · Internet of Things · Smart bins · Route optimization

M. S. Pal (B) · M. Bhatia Lovely Professional University, Phagwara, Punjab, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_58

795

796

M. S. Pal and M. Bhatia

1 Introduction In the digital space, the Internet of Things (IoT) has witnessed a major shift in the last couple of years and it would continue to evolve itself. The IoT is the next frontier with wireless networks, smart sensors and through-going computing capabilities. The idea of IoT is just not about connecting the equipment to the internet altogether but also about sensing the “things” in a smart and real way. The basic principle of IoT is anything, anytime, anywhere, and any network [1, 2]. There is huge growth in the IoT sector and its noteworthy impact on our day-to-day life. Waste management is a core area that is addressed through the IoT in smart cities [3]. According to the studies done globally, it is estimated that the waste produced in the cities will be around 3.5 billion tons by the year 2050. Also, the cost of management of this kind of waste will be around 640 billion dollars which is a huge amount as per the current scenario and the economic situation in this world. There is no doubt that this problem is not going to solve itself and it is a big worry for modern cities all around the world many innovative ways need to be designed to ensure that this problem is solved. Traditional and unscientific ways of Municipal Solid Waste Management (MSWM) impact the environment and hence causes various health hazards to inhabitants [4, 5]. An inadequate solution to waste collection and poor transportation is liable for the accretion of municipal solid waste at every corner and junction [6]. To counter the problems, only the solutions which are related to it in the management domain are not enough and technological solutions for efficient systems are needed, which can enhance the overall efficiency of the waste management system in cities [7, 8]. Since it is necessary to monitor the fill level of waste inside the garbage bins to avoid overflow of bins which creates an unhygienic environment and hence adverse impact on human health. Therefore, smart bins equipped with sensors are being used to monitor the level of garbage inside the smart bins. In a smart city, for the solid waste management system (SWMS) such IoT-enabled smart bins are placed in various places to monitor the real-time status of the fill level of bins. By using a wireless network IoT enabled bins shall capable of transmitting real-time alerts to the command and control center as and when smart bins will be reaching at threshold or full level. The fill level of the bin is detected using ultrasonic sensing. By monitoring the amount of time that passes between delivering and receiving an ultrasonic pulse, the sensor can calculate the distance to a target. When an object is present, the sensor receives a bounced-back ultrasonic pulse that was originally sent out into the air. The distance is calculated using the travel time and sound speed. The garbage collector vehicle will collect garbage from smart bins for treatment/disposal at the dumping site. Figure 1 depicts an end-to-end solution for an SWMS. Without enhancing the routing of garbage collection trucks, we cannot assume the effectiveness of the SWMS [9]. The Optimized and shortest route for collecting waste is predicted by using various algorithms such as greedy algorithms, Digikshitra algorithms, and traveling salesman problems. The Optimized route is based on various parameters such as the number of bins, filling rate, current status, and distance, which depends upon the city to city. Further, IoT-enabled solutions such as Lower-Power Wide Area Network (LPWAN)

Smart Solid Waste Management System Using IoT Technology: …

797

Fig. 1 End-to-end solid waste management system

leveraged low power means long battery life for the device, low bandwidth, and long range could excel as SWMS in smart cities [10]. Hence, this paper brings together the different techniques used for the SWMS with smart IoT-enabled bins by performing reviews on existing research that catalysis end-to-end solutions for the SWMS.

1.1 Paper Organization The paper initiated a basic introduction to IoT and the SWMS. The rest of the paper is outlined in various sections. Section 2, the paper provides a brief overview of the relative and the previous work followed by the comparison of various techniques and algorithms. Section 3, describes various challenges and issues for the Smart Waste Management system. Section 4, covers critical discussion. Finally, Sect. 5 concludes all the review-based research that has been carried out and, the future direction thereof.

2 Analysis of Previous Works As IoT proves a trustworthy one, for SWMS by resolving the issues like accumulating data, processing it, and outputting the result by effectively using certain communication protocols. Data is collected from the real-time sensors. In the data processing phase, data mining and classification techniques are used to process the data into

798

M. S. Pal and M. Bhatia

useful information. The output unit is used to display the results. This section deals with previous works that emphasize IoT-enabled solutions with various routing techniques used for the smart waste management system. The series of recent studies have been categorized in the following three subsections.

2.1 IoT-Enabled Smart Bin The authors in [11], proposed a solution by implementing the sensor-based bins in which the ultrasonic sensors are used to measure the emptiness of the bin and use the LoRaWAN protocol for data collection. The conducted study was a pilot study with outdoor bins and the results show that the existing technology needs improvement to implement this kind of system with inexpensive equipment. This system was not an actual implementation but a prototype to explore the options of placing smart IoT-enabled bins at the university campus. This proposed system did not use any particular algorithm but was just set up to test this system in a realtime environment. The authors presented [12] a system that assists the user in the correct way of sorting and disposing of waste and also conveyed information about the content of each bin. This model has two phases: In the first phase, a smart bin automatically computes the weight of garbage, changes it into reward points, and credits those points into the card. The second phase focus to present the gift to peoples, who abandon their garbage correctly. The gathered points can be exchanged for an item or even withdrawn by the banking system in form of virtual currency. A solution is suggested by the authors [13] to the problem of garbage containers that are not cleaned timely and resultant overflow. The system is embedded with an alarm that triggers when a garbage container is nearby to fill. Further bins are associated with the NIR (near-infrared spectroscopy), a screening system, that enables the identification of the different five types of plastic resin, which are non-biodegradable. Using NIR reflectance spectroscopy to identify and classify plastic resin has many benefits. There are particular NIR wavelengths used to identify the different Plastic resins. Hence, biodegradable waste is easily classified and sent to the production of biogas which is one of the best methods to dispose of the waste. This paper [14] shows that every time information is collected on the use of garbage bins it helps the administrator to detect and decide if a particular place needs extra dumps or to move them to the other required areas. The contractor/bin provider could plan in a better way from daily updated information that when and where need to send a garbage collector vehicle for the emptiness of the filled bins. A “Smart bin” solution is suggested where smart bins are distributed in different places with a unique ID [15]. When bins are about to fill the bin ID and location are communicated through GSM (global system for mobile communication) to the concerned person who is responsible to clean that area.

Smart Solid Waste Management System Using IoT Technology: …

799

2.2 Smart Waste Management in the Context of Smart Cities The research paper [16], focused on the smart management of waste using the IoT and an optimized architecture to manage the waste in the perspective of smart cities where the particular situation of access waste is prevailing. The particular study delved into the quality of typology of the sensor node, based on the cost and the power consumption has to be minimum. The sensing node used in the system consists of a small microcontroller, a sensor used to check the filling up of the status of the bin, and a long-range transmission device based on the LoRa LPWAN technology to look after the overall waste collection process. LoRa LPWAN is a sort of wireless wide-area network created to enable long-distance communications at a low bit rate among the connected things such as sensors operated on a battery. The paper especially analyzed the architecture of the nodes, energy savings during the transmission, and the policies to increase the network longevity. The main purpose was to increase the battery life because the sensors being wireless, need to be energy conserved through hardware and software optimization to increase long working time. The study presented in the paper [17], focused on the Governments and corporations which are looking for solutions to increase the efficiency of garbage collection in cities through the latest technologies. The things used are smart sensors with the IoT and other cloud platforms etc. The proposed method is used to turn the automated garbage collection mechanism into a smart system so that it can be easily included in any smart city setup. Intelligent monitoring is proposed [18] in two phases: the first phase is to monitor the level of waste in the smart bin regularly and in the second phase transportation of the waste through the optimal route. All simulation work is done by the smart-M2 platform, which is an extension of cross-domain search, and through this, it is possible to interoperate applications from different areas. This paper [19] described a cloud-based solution for the waste management system, in which the current status of dustbins is sent to the cloud so that the stockholders may fetch relevant information according to their interests. This solution provides a smart way of handling and disposal waste. There are various intelligent approaches for waste collection and disposal [20]; which can be applied to smart waste management and hence beneficial in terms of cost and environmentally [21, 22].

2.3 Smart Garbage Monitoring and Route Optimizing System Various routing algorithms are used for route optimization of the garbage collection vehicles such as the Mathematical approach, GIS-based, and combined optimized approach. Authors in their work [23], proposed an IoT-based smart garbage management system for the city where the daily task of collecting the generated garbage has to be done. Huge resources are committed to this mechanism but the results are still too vague. Recently we are moving for hybrid approaches with a minimum cost of IoT architecture which presents a better outcome. This paper presented an effective

800

M. S. Pal and M. Bhatia

way for waste management by predicting the filled level of bins. The optimized and shortest route for collecting the waste is predicted by using machine learning and graph theory. The proposed system examined the data transfer on long-range transmission sensors such as the LoRa module and was implemented at Ton Duc Thang University (Vietnam), the results show the system saves lots of time by finding the shortest or best path for waste collection. This paper [24] presented a smart bin for monitoring the garbage and dynamic scheduling for a vehicle to collect the garbage in a smart city. In this approach status of the bin along with the location is sent to the cloud platform. The proposed method also leverages the facility on mobile for the driver of the garbage collector vehicle to follow the shortest path from his present location to the filled bin. The visualization of the gathered data is shown on the ThinkSpeak platform. Data can be collected, visualise, and analyse in real-time data streams in the cloud with ThingSpeak. It is simple to set up devices to communicate with ThingSpeak using well-liked IoT protocols. The author suggested making this system eco-friendly by installing a solar panel for energy to devices which is a truly renewable energy source. The author [25], focused on the retrieved nature of the senses and intentionally focused on the possibility collection algorithm works in the direction of optimizing the waste collection process to automatically retrieve bins with the help of the IoT and the optimized approach for monitoring. Various comprehensive simulations during the study prove that the algorithm proposed is better in terms of quality and quantity criteria which are adapted to analyze the strength and weaknesses of the particular algorithm that is proposed in this work. This research paper [26], addressed that waste management is a multidisciplinary activity that contains the generation, collection, storage, and transportation of waste. The latest technologies and diffusion of the internet together with the advent of compact hardware has allowed the development of efficient waste management system. This study moves towards the coupling of waste collection with the help of GIS web-oriented systems. This article [27], described a different mechanism developed to calculate the efficiency and viability of the system implementation for waste management. A prototype is proposed for wireless nodes using LoRa-WAN to sense the temperature, weight, and level of the bins. The gathered information is analyzed and generates the dynamic route. The result of the case study based on the region of Salamanca shows that the optimized route provides minimum cost, time, and workforce as compared to a static waste collection route. This study shows a developed node has a great operational lifespan, cost, and long coverage with the installation of antennas in the specified region. Another inline procedure in the paper [28] utilizes the Particle Swarm Optimisation which is incorporated with the capacitated vehicle routing problem to determine the best possible route for the waste collection and also make sure that the efficiency is above 75% all the time. Authors in the paper [29] proposed an innovative system that efficiently collects the waste, detects the fire in the waste, and also predicts waste generation. The IoT-enabled devices perform the monitoring of the smart bin and the wireless devices are used to send the status of the filling level of the bins along with its location to the cloud for processing. Two predictive analytic techniques such as neural networks and decision trees are

Smart Solid Waste Management System Using IoT Technology: …

801

used to predict the future generated waste. The proposed system is also enabled for fire detection which is very useful to save human life and economic loss. This study focused on the cost and efficiency of the system. The authors [30] explained modern traceability devices such as volumetric sensors, RFID (Radio Frequency Identification) system, GPRS (General Packet Radio Service), and GPS (Global Positioning System) technology are used to receive the real-time data that is basic to develop an effective routing model for garbage collection. The proposed work was initially implemented in the Italian City which has a hundred thousand inhabitants. This model has been tested and validated by using a simulation tool, which shows the model is economically feasible. The authors in this paper [31], consider a real-life waste collection problem that can be seen as a variant of the vehicle routing problem with time windows. In this approach, a definite time window has been allotted for serving of client, disposal facility, depot, lunch break for driver, and so forth. The study offers an algorithm based on the Ant Colony Meta-heuristics for real-life waste collection problems with better results. The power protocols 6LoWPAN (Pv6 over Low-Power Wireless Personal Area Networks) and RPL (Routing Protocol for LowPower and Lossy Networks) are used for large geographical areas. If the default routes become unreachable, RPL is very adaptive to changing network circumstances by offering alternate routes. RPL performs better than 6LoWPAN when the node density is larger [32].

2.4 Comparative Analysis and Research Gaps A Series of literature have been examined in this paper specially focused on the physical infrastructure of waste bins using IoT technology for an efficient waste management system. The comparative evaluation of this study is depicted in Table 1. In light of the physical infrastructure of IoT enabled bins for smart waste management system has been analysed based on various parameters such as type of waste supported; location of bins (outside or inside); pneumatic pipes tube that automatically compresses garbage to decrease volume; recycling points and processing points for discarded waste to convert in new items or re-processing for correct disposal; types of sensors; GPS and automatic actuators for preventing excessive deposit. Although plenty of techniques have been reported in the literature for handling waste management, however, a closer look reveals that few of the literature considered important features like pneumatic pipes, recycling, and processing point for organic waste, which are the major challenges to dispose of the waste efficiently. After a detailed study, several techniques [33, 34] for waste collection include technology like the IoT and real-time monitoring of smart bins using wireless sensors [35, 36]. It is important to understand that a balance has to be maintained between the energy consumption and the efficiency of the system because the sensors which are working on the bins are wireless and battery-based [37–40]. The energy must be used optimally, otherwise, the sensors are going to die very soon because they will run out of their batteries [41]. Further, the capabilities of a wide number of sensors distributed

802

M. S. Pal and M. Bhatia

in large geographic areas with extended coverage using less power consumption must be taken into consideration. On the other hand, the route selection mechanism must be working most efficiently to ensure that the route selection is the best possible through multiple ways of optimization techniques. A comparative analysis of the smart waste management system is depicted in Table 2. In this review, we have seen a variety of architectures that are being used in which some have real-time data and others have real-time updates through a prototype. In an ideal scenario, the real-time data has to be maintained and interpreted in the prototype, to check the universal approach of the system. Along with it, the route selection has to be efficient through machine learning or with other optimization techniques that can ensure that the locations of the bins are also taken into account for route selection in an optimized manner. The length of the route is optimized to ensure the cost and fuel consumption is minimized. Another ideal aspect that is needed for the smooth mechanism is the long-range transmission option within the nodes along with low power consumption to save energy.

3 Challenges and Issues for Smart Waste Management System Major challenges and issues of the waste management system are described as under.

3.1 System Segregation at the Collection Level While implementing an SWMS, the householders need to augment recycling by themself, which will not only reduce the quantity of waste but also better the performance of the SWMS. The collection and segregation could be done at the source level by rag-pickers, so that reuse or reprocessing could be done efficiently. Thus this practice will reduce the required landfill area and also economically better.

3.2 Optimal Collection Route Out of the total cost of SWMS, 80% of the cost spend on the collection of garbage. The garbage vehicle moves daily to cities for collecting the garbage discarded by the civilians. The main challenge is to provide the shortest and congestion-free path to the garbage vehicle which leads to saving time and money. Hence there is a need to optimize the route for garbage collecting vehicles to collect the filled bins from the different places in the city. The objective of this parameter is to achieve an effective path for garbage collector vehicles in terms of the shortest route, less fuel as well as less time-consuming.

Smart Solid Waste Management System Using IoT Technology: …

803

Table 1 Comparative analysis of smart waste management system Refs. Type of waste

Bin location

Sensors

Pneumatic Recycling Processing GPS Actuators pipes points points

[12]

Glass, plastic, paper, metal

Outside

Capacity

No

Yes

No

No

No

[13]

Plastic

Outside

Capacity, weight

No

Yes

No

Yes

No

[14]

General waste

Outside

Capacity

No

No

No

Yes

No

[15]

General waste

Outside

Capacity

No

Yes

No

No

No

[18]

Glass, plastic, paper, general waste

Outside

Capacity, weight

No

No

No

No

No

[19]

Organic, Underground Capacity glass, plastic, paper, metal

No

Yes

No

No

No

[20]

General waste

Outside

Capacity

Yes

No

No

No

No

[21]

General waste

Outside

Capacity

No

No

No

No

No

[22]

Glass, plastic, paper, metal

Outside

Capacity

No

Yes

No

No

No

[33]

Organic, Outside glass, plastic, paper, metal, toxic

Temperature, Yes humidity, chemical, pressure

Yes

Yes

Yes

Yes

3.3 Communication Technology Various communication technologies and protocols are being used in the IoT applications for sending/receiving the signals/data from the sensor to the network and vice-versa. Some of the prominent technologies are Bluetooth, wi-fi, zig-bee so on so forth. Everyone has its pros and cons. The major challenge with these techniques is short-range, maximum power consumption, and privacy. Hence there is a need to focus on improving their short-range because there are many IoT applications that need the long-range transmission of data/signals. Further, there is also a need

Volume sensor, Different weight sensor heuristics and local search method

HC-SR04 Ultrasonic sensor

Ultrasonic sensor, load sensor

Ultrasonic, flame, weight, and temperature sensor

Ultrasonic sensor

[23]

[24]

[28]

[29]

[33]

–

–

Particle swarm optimization

Google Map API

Neighborhood search algorithm

Ultrasound sensor

[16]

Algorithm

Sensors enabled in smart bins

Refs.

–

–

–

GPS

GIS

–

Use of GIS tool

ThinkSpeak

MATLAB

MATLAB

ThinkSpeak

–

–

Simulation tool

Real-time

Real-data

Real-data

Real-time

Real-data

Real-time

Architecture

No

No

–

Yes

Yes

No

Route selection

Table 2 Literature analysis of IoT-enabled technology used for smart waste management system

Wi-Fi

GPRS

Zigbee

Wi-Fi

Lora WAN

Lora LP-WAN

Communication technology

Published year

Wi-fi is a small range of technology

Unauthorized persons can access the network

Zigbee is prone to attack from unauthorized persons

Wi-fi is a small range of technology

To optimize the techniques to increase the life of the network

(continued)

2019

2020

2020

2018

2018

To improve the 2018 efficiency of the nodes

Limitations

804 M. S. Pal and M. Bhatia

Sensors enabled in smart bins

Ultrasonic sensor, IR sensor

Ultrasonic sensor

Ultrasonic sensor

Weight, ultrasonic sensor

Load sensor, IR sensor

Refs.

[34]

[35]

[36]

[37]

[38]

Table 2 (continued)

–

–

Integer linear programming

Meta-heuristic: backtracking search algorithm

–

Algorithm

–

GIS server

GIS used

–

Ethernet

Use of GIS tool

–

–

Net2Plan

MATLAB

–

Simulation tool

–

–

Real-data

Real-data

Real-Time

Architecture

No

–

Yes

Yes

No

Route selection

Wi-Fi

RFID

GPRS

–

Machine learning concept

Communication technology

Published year

WiFi is short-range communication technology

Needs a person to always have the RFID card upon using the bins for notification

Unauthorized persons can access the network

(continued)

2020

2017

2019

Algorithms do 2017 not work efficiently when more constraints are added

Bins 2017 notification, do not have locations details and ID

Limitations

Smart Solid Waste Management System Using IoT Technology: … 805

Sensors enabled in smart bins

Ultrasonic sensor, weight sensor, gas sensor

Capacity sensor

Weight, temperature, humidity and, Ul

Refs.

[39]

[40]

[41]

Table 2 (continued)

–

ROS and RRT

Heuristic

Algorithm

GPS model (modelNeo-6M)

GPS

–

Use of GIS tool

–

–

–

Simulation tool

Real-time

Real-time

Real-data

Architecture

–

Yes

Yes

Route selection

SIM900 GSM/GPRS module

Lora

SIM900 GPRS, Wi-Fi

Communication technology

Anybody can have access the system because the new user easily register himself

Easily accessible by unauthorized persons

Unauthorized persons can access the network

Limitations

2020

2020

2020

Published year

806 M. S. Pal and M. Bhatia

Smart Solid Waste Management System Using IoT Technology: …

807

to emphasize innovation on less power consumption as well as security for a better system.

3.4 User-Friendly System The waste management system must be people’s centric where they can easily interact with the system through a city App on a mobile platform. Which should have the capacity to receive complaints from various modes like Facebook, Twitter, etc., and dispose of them on the citizen satisfaction, to increase the happiness index of the citizens of the particular city. The city App should also be capable to upload an image of the discarded garbage thrown in the open area, to closely monitor and interact with the general public.

3.5 Lifetime Maximization of IoT Devices An absolute necessity at this point of COVID-19 Lockdown is the automation of legacy the solid management system with the use of the latest entrant technology. The use of the IoT in the SWMS has changed the traditional methods with the latest digital technology and is also helpful in reducing costs. All the IoT-enabled bins in the city are being monitored centrally. But there is also the added benefit of using such technology which leverages end-to-end solutions from collecting to disposal of SWMS. Though IoT-enabled devices play a vital role to improve efficiency such devices are battery-enabled and have limited energy. We can control power transmission, find energy leaks, reduce power during peak hours, and do other things with a power management system. It is affordable to install an energy management system. Therefore need for the hour is to design and develop a holistic technique for IoT-enabled devices which consume less power, to maximize network lifetime in IoT applications for an efficient system. IoT sensors are battery-powered and have limited energy, which reduces the effectiveness of IoT networks as a whole. According to the adaptive sampling and sleep and wake-up technique, each IoT sensor selects an operational mode—such as transmission, sleep, or listening; during each cycle of the garbage collection path based on the filling rate and present condition of the bins. As a result, nodes are built to conserve as much energy as possible in order to extend the lifespan of IoT networks.

4 Critical Discussion The objective of the paper is to study and analyze the state-of-the-art techniques used for the waste management system with smart IoT-enabled bins. This study is

808

M. S. Pal and M. Bhatia

a comprehensive review of various techniques being used for the SWMS by using the IoT. Based on literature analysis, detailed information is typically presented in Table 2. In this table, we have studied mainly two categories of papers is IoT enabled bins and additionally, IoT-enabled bins along with routing algorithms used for the garbage collector vehicle. It has been found there is no significant difference in the overall performance of the system while using IoT-enabled bins without using the optimizing routing algorithm. By and large, the communication tools used to send the status of the bin to the receiving end are short-ranged and sometimes not able to send the full information of the smart bin. Further, there is also a need to emphasize the physical infrastructure of waste bins using IoT for an efficient waste management system. The article makes no mention of waste policies to be improved in order to have a sustainable garbage collection system in the future.

5 Conclusion It has been observed that SWMS is a prominent area that has to be sorted out efficiently in every city. In day to day delightful acceptance of IoT, most of the connected edge devices around us have penetrated more or less every aspect of life. The absence of IoT platforms in the SWMS has manifestly been formidable in terms of everything. Hence, IoT is redefining the way SWMS was looked at initially. This paper analyzed the smart waste management system using the IoT and also route optimization methods used by the garbage vehicle. Apart from this, the paper also highlighted the various challenges and issues of the smart waste management system. Hence, it is significant that invention and enrichment be geared to manage the waste in our smart cities to ensure and sustain the foremost quality of life with a healthy ecosystem. As a future scope, it is suggested that improvement in communication technology which leads to increased efficiency and extended coverage using less power consumption and reliability may be used with IoT-enabled solutions for the effectiveness of SWMS. Future research could focus on extending the solar-powered IoT sensor to enhance the life expectancy of sensing nodes and use unmanned aerial vehicles, or “drones,” for garbage collection and monitoring. Future research challenges may focus further on vehicle routing for garbage collection with territorial-specific parameters. Furthermore, the route optimization techniques used in solid management need further enrichment for fuel efficiency and crucial cost. The physical infrastructure of garbage bins using IoT which supports various parameters may be further improved for a better management system. Thus, using all the above features will certainly enhance the overall SWMS.

Smart Solid Waste Management System Using IoT Technology: …

809

References 1. Gubbi J, Buyya R, Marusic S, Pal Aniswami M (2013) Internet of things (iot): a vision, architectural elements, and future directions. Future Gener Comput Syst 29(7):1645–1660 2. Bhatia M, Sood SK (2018) Internet of things based activity surveillance of defence personnel. J Ambient Intell Humanized Comput 9(6):2061–2076 3. Srinidhi NN, Dilip Kumar SM, Venugopal KR (2019) Network optimizations in the internet of things: a review. Eng Sci Technol Int J 22(1):1–21 4. Minghua Z, Xiumin F, Rovetta A, Qichang H, Vicentini F, Bingkai L, Giusti A, Yi L (2009) Municipal solid waste management in Pudong new area, China. Waste Manage 29(3):1227– 1233 5. Rathi S (2006) Alternative approaches for better municipal solid waste management in Mumbai, India. Waste Manage 26(10):1192–1200 6. Kaushal RK, Varghese GK, Chabukdhara M (2012) Municipal solid waste management in India-current state and future challenges: a review. Int J Eng Sci Technol 4(4):1473–1489 7. Schaffers H, Komninos N, Pallot M, Trousse B, Nilsson M, Oliveira A (2011) Smart cities and the future internet: towards cooperation framework for open innovation. In: The future internet assembly. Springer, Berlin, Heidelberg, pp 431–446 8. Pavithra D, Balakrishnan R (2015) Iot based monitoring and control system for home automation. In: 2015 global conference on communication technologies (GCCT). IEEE, pp 169–173 9. Tirkolaee EB, Mahdavi I, Esfahani MMS, Weber G-W (2020) A hybrid augmented ant colony optimization for the multi-trip capacitated arc routing problem under fuzzy demands for urban solid waste management. Waste Manage Res 38(2):156–172 10. Mdukaza S, Isong B, Dladlu N, Abu-Mahfouz AM (2018) Analysis of iot-enabled solutions in smart waste management. In: IECON 2018–44th annual conference of the IEEE industrial electronics society. IEEE, pp 4639–4644 11. Lundin AC, Ozkil AG, Schuldt-Jensen J (2017) Smart cities: a case study in waste monitoring and management. In: Proceedings of the 50th Hawaii international conference on system sciences 12. Abd Wahab MH, Kadir AA, Tomari MR, Jabbar MH (2014) Smart recycle bin: a conceptual approach of smart waste management with integrated web based system. In: 2014 international conference on IT convergence and security (ICITCS). IEEE, pp 1–4 13. Thakker S, Narayanamoorthi R (2015) Smart and wireless waste management. In 2015 international conference on innovations in information, embedded and communication systems (ICIIECS). IEEE, pp 1–4 14. Folianto F, Low YS, Yeow WL (2015) Smartbin: smart waste management system. In: 2015 IEEE tenth international conference on intelligent sensors, sensor networks and information processing (ISSNIP). IEEE, pp 1–2 15. Ramya E, Sasikumar R (2017) A survey of smart environment conservation and protection for waste management. In: 2017 third international conference on advances in electrical, electronics, information, communication and bio-informatics (AEEICB). IEEE, pp 242–245 16. Cerchecci M, Luti F, Mecocci A, Parrino S, Peruzzi G, Pozzebon A (2018) A low power iot sensor node architecture for waste management within smart cities context. Sensors 18(4):1282 17. Popa CL, Carutasu G, Cotet CE, Carutasu NL, Dobrescu T (2017) Smart city platform development for an automated waste collection system. Sustainability 9(11):2064 18. Catania V, Ventura D (2014) An approach for monitoring and smart planning of urban solid waste management using smart-m3 platform. In: Proceedings of 15th conference of open innovations association FRUCT. IEEE, pp 24–31 19. Aazam M, St-Hilaire M, Lung C-H, Lambadaris I (2016) Cloud-based smart waste management for smart cities. In: 2016 IEEE 21st international workshop on computer aided modelling and design of communication links and networks (CAMAD). IEEE, pp 188–193

810

M. S. Pal and M. Bhatia

20. Saha HN, Auddy S, Pal S, Kumar S, Pandey S, Singh R, Singh AK, Banerjee S, Ghosh D, Saha S (2017) Waste management using internet of things (iot). In: 2017 8th annual industrial automation and electromechanical engineering conference (IEMECON). IEEE, pp 359–363 21. Chowdhury B, Chowdhury MU (2007) Rfidbased real-time smart waste management system. In: 2007 Australasian telecommunication networks and applications conference. IEEE, pp 175–180 22. Lu J-W, Chang N-B, Liao L, Liao M-Y (2015) Smart and green urban solid waste collection systems: advances, challenges, and perspectives. IEEE Syst J 11(4):2804–2817 23. Khoa TA, Phuc CH, Lam PD, Bao Nhu LM, Trong NM, Thi N, Phuong H, Van Dung N, Tan-Y N, Nguyen HN, Minh Duc DN (2020) Waste management system using iot-based machine learning in university. Wireless Commun Mob Comput 24. Chaudhari SS, Bhole VY (2018) Solid waste collection as a service using iot-solution for smart cities. In: 2018 international conference on smart city and emerging technology (ICSCET). IEEE, pp 1–5 25. Anagnostopoulos T, Kolomvatsos K, Anagnostopoulos C, Zaslavsky A, Hadjiefthymiades S (2015) Assessing dynamic models for high priority waste collection in smart cities. J Syst Softw 110:178–192 26. Rada EC, Grigoriu M, Ragazzi M, Fedrizzi P (2010) Web oriented technologies and equipments for msw collection. Proceedings of the International Conference on Risk Management, Assessment and Mitigation-RIMA 10:150–153 27. Lozano A, Caridad J, De Paz JF, Gonzalez GV, Bajo J (2018) Smart waste collection system with low consumption Lorawan nodes and route optimization. Sensors 18(5):1465 28. Hannan MA, Akhtar M, Begum RA, Basri H, Hussain A, Scavino E (2018) Capacitated vehicle routing problem model for scheduled solid waste collection and route optimization using pso algorithm. Waste Manage 71:31–41 29. Jim AAJ, Rafiul Kadir M, Mamun AA, Abdullah-Al Nahid M, Ali et al (2019) A noble proposal for internet of garbage bins (iogb). Smart Cities 2(2):214–229 30. Faccio M, Persona A, Zanin G (2011) Waste collection multi objective model with real time traceability data. Waste Manage 31(12):2391–2405 31. Islam R, Sohel Rahman M (2012) An ant colony optimization algorithm for waste collection vehicle routing with time windows, driver rest period and multiple disposal facilities. In: 2012 international conference on informatics, electronics & vision (ICIEV). IEEE, pp 774–779 32. Hariharakrishnan J, Bhalaji N (2021) Adaptability analysis of 6LoWPAN and RPL for healthcare applications of ınternet-of-things. J ISMAC 3(02):69–81. https://doi.org/10.36548/jismac. 2021.2.001 33. Atayero AA, Williams R, Badejo JA, Popoola SI (2019) Cloud based iot-enabled solid waste monitoring system for smart and connected communities. Int J Civil Eng Technol 10(2):2308– 2315 34. Baby CJ, Singh H, Srivastava A, Dhawan R, Mahalakshmi P (2017) Smart bin: an intelligent waste alert and prediction system using machine learning approach. In: 2017 international conference on wireless communications, signal processing and networking (WiSPNET). IEEE, pp 771–774 35. Akhtar M, Hannan MA, Begum RA, Basri H, Scavino E (2017) Backtracking search algorithm in cvrp models for efficient solid waste collection and route optimization. Waste Manag 61:117– 128 36. Bueno-Delgado M-V, Romero-Gázquez J-L, Jiménez P, Pavón-Marino P (2019) Optimal path planning for selective waste collection in smart cities. Sensors 19(9):1973 37. Al-Jabi M, Diab M (2017) Iot-enabled citizen attractive waste management system. In: 2017 2nd international conference on the applications of information technology in developing renewable energy processes & systems (IT-DREPS). IEEE, pp 1–5 38. Ali T, Irfan M, Alwadie AS, Glowacz A (2020) Iot-based smart waste bin monitoring and municipal solid waste management system for smart cities. Arab J Sci Eng 45:10185–10198 39. Johnson M, Emilin Shyni C Smart garbage bin with efficient routing and management system 14:80–89

Smart Solid Waste Management System Using IoT Technology: …

811

40. Zhang Q, Li H, Wan X, Skitmore M, Sun H (2020) An intelligent waste removal system for smarter communities. Sustainability 12(17):6829 41. Anagnostopoulos T, Zaslavsky A, Kolomvatsos K, Medvedev A, Amirian P, Morley J, Hadjieftymiades S (2017) Challenges and opportunities of waste management in iot-enabled smart cities: a survey. IEEE Trans Sustain Comput 2(3):275–289

HLWEA-IOT: Hybrid Lightweight Encryption Algorithm Based Secure Data Transmission in IoT-MQTT Networks S. Hariprasad, T. Deepa, and N. Bharathiraja

Abstract Internet of things (IoT) devices can store and manage the real-time data created by many restricted Internet-connected devices. If one of the nodes were compromised due to Man-in-the-Middle (MITM) attack, the network might suffer significant damage. Due to the limited resources of constrained devices, it is difficult to incorporate appropriate cryptographic capabilities. Hence lightweight cryptography strives to meet the security needs of situations with few resource-constrained devices. In this paper, the framework is constructed using the smart aircraft environment monitoring system (SAEMS) and created with the help of nodes and the message queuing telemetry transport (MQTT) protocol for communicating the sensor data. A hybrid lightweight encryption algorithm (HLWEA) is proposed to mitigate the MITM attack on IoT devices. The HLWEA comprises (i) Key generation and (ii) encryption and decryption. The proposed method achieves an encryption time of 0.0309 ms; encryption bandwidth is 19.02 kbps, decryption time of 0.029 ms and decryption bandwidth of 19.36 kpbs. The proposed implementation is a smaller key size, minimal time complexity, and enhanced real-time cryptography-capable security. Keywords Internet of things (IoT) · Man-In-The-Middle attack · Lightweight cryptography · Key generation · Light encryption device

S. Hariprasad (B) · T. Deepa SRM Institute of Science and Technology, Kattankulathur, Tamilnadu, India e-mail: [email protected] T. Deepa e-mail: [email protected] N. Bharathiraja Dept. of Bio Medical Engineering, Vel Tech Multi Tech Dr. Rangarajan Dr. Sakunthala Engineering College, Avadi, India © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_59

813

814

S. Hariprasad et al.

1 Introduction Internet of Things (IoT) devices are becoming more prominent in many domains which acquire and exchange the data connected to the internet [20]. IoT components are everything: any device, anybody, any service, company, path, network, anytime, and link. According to Statista [25], 30.9 billion IoT devices will be linked worldwide in 2025. Connectivity across infrastructure and services is not possible now without the development of IoT. The three levels of security aspects in IoT such as (1) Design security, (2) Hardware security (3) Data security. Data security has a lot of problems, like dealing with huge amounts of data and keeping track of all the IoT devices. For more extensive data, more time-consuming security algorithms like Advanced Encryption Standards (AES) [18] and Data Encryption Standards (DES) [21] are needed to protect the data in IoT networks. Therefore, lightweight cryptography can support a smaller number of bits in the maximum of 32 or 64 bits of data with smaller keys up to 64 bits. There are other limitations on lightweight algorithms, which need to be looked into during the first phase of the standardization process. Lightweight cryptography is classified into block ciphers, hash functions, message authentication codes, and stream ciphers as shown in Fig. 1. Through a series of rounds, a block cipher encrypts the data bit by bit. These block cipher structures are classified into further types, such as substitution permutation network (SPN) and Feistel Network. In-Stream cipher encrypts one or two bits of data at a time. A hash function is a function that randomly encrypts the data according to the length of the data. The block cipher-based lightweight cryptographic primitives have performance advantages over other cryptographic standards based on power, energy consumption, latency and throughput. Therefore lightweight cryptography can support a smaller number of bits maximum of 32 or 64 bits of data with smaller keys up to 64 bits. The detailed algorithms/methods of all block cipher based lightweight cryptography and types of structure of block ciphers were discussed in the Table 1.

Fig. 1 Different types of lightweight cryptography methods

HLWEA-IOT: Hybrid Lightweight Encryption Algorithm Based … Table 1 Works related to lightweight cryptography References Method [4] [29] [14] [27] [3] [10] [6] [17] [23] [5]

PRESENT (symmetric) RECTANGLE (symmetric) HIGHT (symmetric) TWINE (symmetric) SIMON (symmetric), SPECK (symmetric) LED (symmetric) KTANTAN32 (symmetric) RSA (asymmetric) ECC (asymmetric) PRINCE (symmetric)

815

Features Ultra-lightweight Fast implementation Ultra-lightweight Energy efficient Key schedule Used in RFID tags Algebric equation More secured High speed Low delay

The main research contribution is described as follows. • A novel framework of simulation setup for a smart aviation environment monitoring system (SAEMS) is created with ten nodes. • A novel HLWEA is formed using (i) Key generation and (ii) Encryption and decryption. • A HLWEA is applied to protect the heterogeneous sensor data for secure transfer. • Finally, an assessment is done to show the performance analysis of the proposed scheme, and it is compared with the existing system metrics such as computational time complexity.

2 Related Works The IoT devices are resource-constrained to reduce the communication overhead. Lightweight protocol called message queuing telemetry transport (MQTT) is used for publishing/subscribing messaging to client-server communication [8]. It is designed as lightweight and easy to implement. These protocols can be used in restricted communications environments in the machine to machine (M2M), and less network bandwidth is needed. The various security aspects, application-related and compared protocols with MQTT are discussed below [19]. An IoT-based decision device for intelligent irrigation systems was developed using two protocols such as MQTT and HTTP. Compared to MQTT, other works of literature used a heavy computation protocol such as HTTP to transfer the sensor data to the cloud. Studies have shown that using IoT with MQTT protocol has a lot of benefits [16]. Smart grids can now transmit and interpret data in real-time. The multi-tier edge computing model was developed [28] using two MQTT as a remote broker in fog and cloud. As part of [22]. MQTT messages are exchanged between several nodes,

816

S. Hariprasad et al.

including MQTT Control Packets. There are three principal parts in an MQTT control packet: the fixed header, the variable header, and the payload. In recent years, edge computing has gained popularity among cloud service providers, such as Amazon, Google, and Microsoft [1], due to its low latency IoT connectivity and large data processing capacity. The architecture of the MQTT protocol is open source and easily vulnerable to attacks. The study [2] eventually led to a more robust protocol development for IoT-related applications. Secure-MQTT [12] uses fuzzy logic to detect malicious MQTT broker node activities. MQTT publisher traffic patterns are taken into consideration while selecting traffic features [9]. The attack simulation results in a network with 10 to 50 % malicious nodes, resulting in many malicious nodes. The usage of the MQTT protocol in IoT with the working of encrypting the message transfers in applications is demonstrated in [11]. Fiestal network structure converts 64-bit input into two 32-bit output. RPP decodes data using logical XOR, encrypts and decrypts data via bit swapping and recursive positional substitution on prime and nonprime of cluster results in 64-bit cipher [7]. Data from many heterogeneous IoT devices with attack detection modelling named the SENMQTTSET has been discussed [24]. A study of alternative media transport mechanisms in IoT networks is conducted [13] using constrained application protocol (CoAP) and MQTT-SN for media propagation in low power lossy networks (LLNs). Order messages and re-send lost messages are essential components of IoT’s reliable message communication system [15]. An algorithm known as Improved Artificial Bee Colony (IABC) [26] is used to determine when the key should be upgraded for maximum effectiveness.

2.1 Problem Statement The nodes in an aviation monitoring system are resource-constrained devices on the heavy computation cryptography algorithm, leading to computational complexity and time decay. The proposed Hybrid Lightweight Encryption Algorithm (HLWEA) method addresses this issue. This HLWEA proves that eavesdropping has been a complication and cannot hijack the nodes.

3 Methodology The proposed methodology comprises two parts (a) creating a simulation setup framework for SAEMS and (b) securing the data using HLWEA.

HLWEA-IOT: Hybrid Lightweight Encryption Algorithm Based …

817

Fig. 2 Framework for smart aviation environment monitoring system (SAEMS)

3.1 Framework of SAEMS The proposed framework consists of four modules such as (i) Sensors, (ii) Threat analysis, (iii) Cloud and (iv) Graphical user interface (Control room), as illustrated in Fig. 2. Sensors The SAEMS framework is constructed by using ten nodes by temperature and humidity sensors. The temperature sensor is used to sense the engine’s temperature, and humidity is used to monitor the environment at an instance of time. These two parameters are vital in preventing any hazards to the aircraft nodes. The sensed data of all nodes is fetched and communicated to the server using the MQTT protocol. The collected sensor information is encrypted using the proposed HLWEA transferred through the MQTT protocol. The secure sensor data is like a publisher, and the data analytics section is like a subscriber. The cloud works as a broker MQTT, checking the publisher and subscriber identity using topics and keys generated. Threats There are more chances to prevail against more threats in the fog layer for the data. These threats could be controlled and precise from the forging of collected data. Sometimes these threats can destroy the devices by node traps like MITM. To overcome the existing threats, HLWEA is proposed for more real-time security without modifying any existing service architecture. Cloud MQTT broker will act as a cloud. It has two other main components, such as publisher and subscriber. The node of the publishing packet needs to be encrypted. The sensed data of all nodes is fetched and communicated to the MQTT broker with a specific topic. The collected sensor information is encrypted and decrypted using the proposed HLWEA.

818

S. Hariprasad et al.

Graphical User Interface(GUI) The collected information from the cloud server is given to the control room for data analytics. The received encrypted data will be decrypted at actuators using a micro python script with the help of the key of each node. The GUI part was created with the help of node-red and MQTT brokers. The collected sensor data from the publisher node is encrypted using the lightweight proposed cryptography method using the HLWEA algorithm. The encrypted data will be decrypted at the GUI admin using python script and can be able to witness the node environment details at every instance.

3.2 Hybrid Lightweight Encryption Algorithm (HLWEA) The HLWEA is used to secure the sensor data of the nodes and transmit it to the receiver for data analytics. The proposed HLWEA system consists of a lightweight cypher. It consists of (i) Key generation and (ii) Encryption/Decryption. The proposed HLWEA is unbreakable and compatible with lightweight devices. The embedded system needs more hardware components, cost, area and power consumption; hence, a new replacement for hardware components is a software aspect called lightweight cryptography.This ciphers are solutions for hardware and more secured transmission. Figure 3 illustrates the proposed flow diagram HLWEA and the steps involved in the proposed method. Initially, the temperature sensor data of the aircraft nodes is considered as plain bits and the key is generated using the SPECK key scheduling algorithm using the MQTT topic, which is a public key. Later, using encryption for encrypting, the plain bits are changed to cipher bits and sent to the end user using the MQTT protocol. In the GUI, the encrypted bits are decrypted with the of a public key of the MQTT topic by using a decryption algorithm. Key generation SPECK Key generation is based on two steps (i) round functions and (ii) Key schedule. In the round function, the sensor data is first added bitwise of the key taken from the MQTT topic. Later by addition modulo and key schedule of left and right circular shift. Finally generated, the key for encryption and decryption to be performed. Encryption and Decryption After the SPECK key schedule with add of sensor data is further encrypted using four steps and 8, 12 rounds. The four steps are (i) Add constant and Shift columns, (ii) Substitute S box, (iii) Shift rows and (iv) Mix Columns. Add Constant and Shift columns The constants are added to the consecutive bits of data received after adding the key. The columns are shifted. Substitute S box For substitute S box, the RECTANGLE method is used. It is also a lightweight block cipher and works fast in implementation. RECTANGLE has a four-bit box substitution. RECTANGLE consumes low power than any other box and is a more secure way of transmission and prevents linear attacks.

HLWEA-IOT: Hybrid Lightweight Encryption Algorithm Based …

819

Fig. 3 Proposed flow diagram for ELWEA

Shift rows The rows are shifted for encrypting using the linear feedback shift keying method to shift the bits. Mix Columns At last, The column is mixed once again for shuffling the bits to get stronger encrypted code. Each constant value is mixed using a linear feedback shift register and updated with a new value.

820

S. Hariprasad et al.

4 Results and Discussions The experimental setup of 10 aircraft nodes has been constructed with the help of python 3.6. Two vital parameters of aircraft nodes, such as temperature and humidity data, are generated randomly at an instance of time. Figure 4 illustrates the attack design framework of the proposed model. The proposed HLWEA protects the data from aircraft nodes to the control room. If an attacker gets an MQTT packet, they can not decrypt the data without the key.

Fig. 4 Attack design framework Fig. 5 Message received to the control room without encryption

HLWEA-IOT: Hybrid Lightweight Encryption Algorithm Based …

821

Fig. 6 Message received to the control room with proposed HLWEA Table 2 Proposed HLWEA Performance metrics Method Encryption Decryption time time (ms) (ms) Proposed

0.0309

0.029

Encryption Throughput (kbps)

Decryption Throughput (kbps)

19.02

19.36

Figures 5 and 6 shows the data transmitted and received between aircraft nodes to the data centre without encryption and the proposed HLWEA algorithm. The proposed encryption algorithm has been deployed in all ten aircraft nodes. Sensor nodes will act as a publisher, MQTT broker servers as a cloud, and the end data centre is called a subscriber. During this time, the publisher sends with and without encryption parts are reflect the subscriber system. Table 2 shows the various comparison like block size and key in terms of bits and time in seconds compared to the proposed method. To calculate the node lifetime of the proposed model, the voltage of each node is considered as 3.3 V DC, and the current is ten microamps. The total lifetime of the nodes is around five years.

5 Conclusion This study describes end-to-end payload MQTT-based IoT devices. Ten nodes using the MQTT protocol make up SAEMS infrastructure to protect data against spoofing, and HLWEA was proposed. Initially the ten nodes are considered without encryption, and then ten nodes are considered with HLWEA. Compared to the other literature,

822

S. Hariprasad et al.

the proposed method achieved less time complexity of encryption time of 0.0309 ms. Encryption bandwidth is 19.02 kbps, decryption time of 0.029 ms and decryption bandwidth of 19.36 kpbs. In future work, the same lightweight encryption can be implemented other protocols.

References 1. Ahmad T, Morelli U, Ranise S, Zannone N (2022) Extending access control in AWS IoT through event-driven functions: an experimental evaluation using a smart lock system. Int J Inf Secur 21(2):379–408 2. Akhtar S, Zahoor E (2021) Formal specification and verification of MQTT protocol in pluscal-2. Wireless Pers Commun 119(2):1589–1606 3. Beaulieu R, Shors D, Smith J, Treatman-Clark S, Weeks B, Wingers L (2015) The simon and speck lightweight block ciphers. In: Proceedings of the 52nd annual design automation conference, pp 1–6 4. Bogdanov A, Knudsen LR, Leander G, Paar C, Poschmann A, Robshaw MJ, Seurin Y, Vikkelsoe C (2007) Present: an ultra-lightweight block cipher. In: International workshop on cryptographic hardware and embedded systems, Springer, pp 450–466 5. Borghoff J, Canteaut A, Güneysu T, Kavun EB, Knezevic M, Knudsen LR, Leander G, Nikov V, Paar C, Rechberger C et al. (2012) Prince–a low-latency block cipher for pervasive computing applications. In: International conference on the theory and application of cryptology and information security, Springer, pp 208–225 6. Cannière CD, Dunkelman O, Kneževi´c M (2009) Katan and ktantan-a family of small and efficient hardware-oriented block ciphers. In: International workshop on cryptographic hardware and embedded systems, Springer, pp 272–288 7. Chatterjee R, Chakraborty R, Mondal J (2019) Design of lightweight cryptographic model for end-to-end encryption in IoT domain. IRO J Sustain Wirel Syst 1(4):215–224 8. Edited by Andrew Banks and Rahul Gupta: MQTT Version 3.1.1. (2014). http://docs.oasisopen.org/mqtt/mqtt/v3.1.1/os/mqtt-v3.1.1-os.html 9. González-Zapata AM, Tlelo-Cuautle E, Cruz-Vega I, León-Salas WD (2021) Synchronization of chaotic artificial neurons and its application to secure image transmission under MQTT for IoT protocol. Nonlinear Dyn 104(4):4581–4600 10. Guo J, Peyrin T, Poschmann A, Robshaw M (2011) The led block cipher. In: International workshop on cryptographic hardware and embedded systems, Springer, pp 326–341 11. Gupta V, Khera S, Turk N (2021) MQTT protocol employing IoT based home safety system with ABE encryption. Multimedia Tools Appl 80(2):2931–2949 12. Haripriya A, Kulothungan K (2019) Secure-MQTT: an efficient fuzzy logic-based approach to detect dos attack in MQTT protocol for internet of things. EURASIP J Wirel Commun Netw 2019(1):1–15 13. Herrero R (2020) MQTT-SN, COAP, and RTP in wireless IoT real-time communications. Multimedia Syst 26(6):643–654 14. Hong D, Sung J, Hong S, Lim J, Lee S, Koo BS, Lee C, Chang D, Lee J, Jeong K et al (2006) Hight: A new block cipher suitable for low-resource device. In: International workshop on cryptographic hardware and embedded systems, Springer, pp 46–59 15. Hwang HC, Park J, Shon JG (2016) Design and implementation of a reliable message transmission system based on MQTT protocol in IoT. Wireless Pers Commun 91(4):1765–1777 16. Kondoro A, Dhaou IB, Tenhunen H, Mvungi N (2021) Real time performance analysis of secure IoT protocols for microgrid communication. Futur Gener Comput Syst 116:1–12 17. Mahto D, Khan DA, Yadav DK (2016) Security analysis of elliptic curve cryptography and RSA. In: Proceedings of the world congress on engineering, vol 1, pp 419–422

HLWEA-IOT: Hybrid Lightweight Encryption Algorithm Based …

823

18. Moradi A, Poschmann A, Ling S, Paar C, Wang H (2011) Pushing the limits: a very compact and a threshold implementation of AES. In: Annual international conference on the theory and applications of cryptographic techniques, Springer, pp 69–88 19. Nawandar NK, Satpute VR (2019) IoT based low cost and intelligent module for smart irrigation system. Comput Electron Agric 162:979–990 20. Ray PP (2018) A survey on internet of things architectures. J King Saud Univ Comput Inf Sci 30(3):291–319 21. Satoh A, Morioka S (2003) Hardware-focused performance comparison for the standard block ciphers AES, camellia, and triple-des. In: International conference on information security, Springer, pp 252–266 22. Seoane V, Garcia-Rubio C, Almenares F, Campo C (2021) Performance evaluation of COAP and MQTT with security support for IoT environments. Comput Netw 197:108338 23. Sharma S, Chopra V (2017) Data encryption using advanced encryption standard with key generation by elliptic curve diffie-hellman. Int J Sec Appl 11(3):17–28 24. Siddharthan H, Deepa T, Chandhar P (2022) SenMOTT-set: an intelligent intrusion detection in IoT-MQTT networks using ensemble multi cascade features. IEEE Access 10:33095–33110 25. Statista Research Department: Internet of Things (IoT) connected devices installed base worldwide from 2015 to 2025 (2016). https://www.statista.com/statistics/471264/iot-number-ofconnected-devices-worldwide/ 26. Suma V, Haoxiang W (2020) Optimal key handover management for enhancing security in mobile network. J Trends Comput Sci Smart Technol (TCSST) 2(04):181–187 27. Suzaki T, Minematsu K, Morioka S, Kobayashi E (2011) Twine: a lightweight, versatile block cipher. In: ECRYPT workshop on lightweight cryptography, vol 2011 28. Veeramanikandan M, Sankaranarayanan S (2019) Publish/subscribe based multi-tier edge computational model in internet of things for latency reduction. J Parallel Distrib Comput 127:18–27 29. Zhang W, Bao Z, Lin D, Rijmen V, Yang B, Verbauwhede I (2015) Rectangle: a bit-slice lightweight block cipher suitable for multiple platforms. Sci China Inf Sci 58(12):1–15

A Practical Approach for Crop Insect Classification and Detection Using Machine Learning Ravindra Yadav and Anita Seth

Abstract Insect identification is one of the most pressing difficulties for Indian farmers, as numerous insect species harm a vast number of crops and hence diminish the quality of harvests, resulting in financial losses for both farmers and the country. However, in agriculture, the combination of IoT and machine learning (ML) allows for ease and innovation, allowing farmers all over the world to better their farming operations. On the other hand, in India, a very little amount of farmers is aware of smart farming and its benefits. Various study on research paper shows that the proper use of IoT devices embedded with the machine learning algorithm can reduce the task of farmer at very early stage of the plant life and thus saving the crops from being degraded, also included the survey of various research done across the globe and identified the potential methods which must be included for the current era farmers in order to minimize the insect effect on the crops. The aim of our experiment is to involve ML and IoT technology to sense the crop conditions in terms of quality and whether it is affected by insect or not for this a experimental study with the help of image processing has been performed thus calculation of results done accordingly. There are various sensors, which are equipped with ML technology like computer vision algorithm, which make the sensor powerful, and images being captured by these sensor can be analysed automatically and thus trigger the automated pesticide treatment systems using a ML-based decision support model. In this paper, study of Convolution Neural Network (CNN), Long Short Term Memory (LSTM), Support Vector Machine (SVM), Grid search based SVM (Grid-SVM), and K-nearest Neighbour classifier has been done Among them based on the required performance, nd the CNN-based model is much accurate for predicting the required treatment. The CNN has achieved up to 88% of accurate classification. Further, the model has been extending by incorporating the regression analysis, which enables the system to R. Yadav (B) · A. Seth IET DAVV, Indore, Madhya Pradesh, India e-mail: [email protected] A. Seth e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 J. Hemanth et al. (eds.), Intelligent Cyber Physical Systems and Internet of Things, Engineering Cyber-Physical Systems and Critical Infrastructures 3, https://doi.org/10.1007/978-3-031-18497-0_60

825

826

R. Yadav and A. Seth

recommend the quantity of the required treatment. In this context, study on the KNN regression and Support Vector Regression (SVR) model has been made, among them the KNN regression provides up to 99.8% accurate prediction for treatment quantity prediction. Keywords Machine learning · Internet of Things · Pest control · Supervised learning · Insect class analysis

1 Introduction Many obstacles can develop in terms of data limitation, uneven data count, and background noise when it comes to automatically identifying the insects harming the crops. For improved performance, they must all be overcome [1]. Mixing CNNs based on numerous topologies (EfficientNetB0, ResNet50, GoogleNet, ShuffleNet, tMobileNetv2, and DenseNet201) and various Adam optimization techniques performed well for pest identification. Two novel Adam algorithms based on the Adam variant DGrad for deep network optimization are proposed, which include a scaling component in the learning rate that is applied to the absolute difference term. On three benchmark bug data sets, CNNs with different data augmentation to assure variety or different types of Adam optimization were trained. Three evaluation measures were used to compare and evaluate fusions. The best performing ensemble, made up of CNN ensembles and the additional variations introduced here, is demonstrated to outperform the literature on all three tests [2]. The highly accurate method for the detection of bug and pest in the farm inspire the farmers for adoption of new technologies to work in their field [3]. When it comes to automatically detecting the insects causing crop damage, many challenges can arise due to data limitations, unequal data counts, and background noise. They must all be conquered in order to increase performance [1]. For pest detection, a combination of CNNs based on a variety of topologies (EfficientNetB0, ResNet50, GoogleNet, ShuffleNet, tMobileNetv2, and DenseNet201) with several Adam tuning approaches performed well. This paper shows, the highlights of recently carried out our work in this area [4–8]. In our recent contribution [2nd paper reference], the IoT-enabled and MLbased technique employment in agriculture has been considered. In this paper, an overview of these contributions hase been discussed in the paper Further discusses the objectives, which were established previously. Further, the experimental study-based identified ML algorithm has used to design a predictive model, which recommends the relevant treatment for the plants. In this context, different deep learning models has been modelled, which will learn on the IoT sensor-based plant health conditions traces and predict the requirement of the pest control [9–12]. The work has also investigated the regression analysis techniques to predict the required quantity of the treatment. Therefore the paper includes the implementation and results analysis of two different architectures of smart farming module. Among them, first one is a low cost solution and can be deploy only with the treatment prediction. The second

A Practical Approach for Crop Insect Classification and Detection Using …

827

model is a fully featured model, use of classification as well as regression technique has been made for providing an exclusive system for automating pesticide treatment [12, 13]. This section provides the overview of the proposed work involved in this paper, the next sections provide the summary of recent contributions, further, In this paper discussion on an experimental study and their consequences has been shown, and then proposed the required smart farming models for the prediction of water, fertilizer and pesticide treatment of the crops. Finally, by using the experimentations, the results have discussed and the future direction of research work has been proposed.

2 Literature Review The recent technologies, methods, and applications in agriculture have taken into consideration to collect 50 research articles from different journals using Google Scholar. These contributions has summarized in this section.

2.1 Noteworthy Contribution The source has given the wide specification and classifies the insects based on their nature, and provides a wide spectrum to understand the kind of diseases associated with a specific type of insects. The below table provide the details. Wu et al. [14] has addressed the problems with agricultural product output caused by the presence of insects in the field, emphasizing the importance of having an accurate dataset for insect identification so that timely preventative measures can be done to avoid economic losses. They gathered a large scale dataset for insect pest recognition called IP102. There are almost 75,000 photos in the dataset, divided into 102 categories. Thenmozhi and Srinivasulu Reddy [15] while identifying the issue of pest detection, the writers of this research took into account a variety of crops. They described how to use a Convolution neural network (CNN) in conjunction with deep architecture to solve the problem of identifying different insects that have very little difference in terms of shape and size among them. The CNN performs automatic feature extraction and learns complex high-level features in image classification applications. Caballero et al. [16] use hyper spectral imaging (HSI) and multispectral imaging (MSI) to address crop health, water or fertiliser use, and probable sickness. Writer has provided a description that gives an overview of some of the pertinent scientific literature on the use of HSI and MSI on agriculture fields. Some of the applications include the detection of contaminants and heavy metals, as well as the management and assessment of water.

828

R. Yadav and A. Seth

Kulkarni et al. [17] proposed to generate messages to notify farmers. That will assist farmers by getting data from the land to take necessary steps to do. Jankielsohn [18] discusses the different varieties of the insect which are friend of the farmer and evaluate a methodology how to increase the growth of good insects, what ae the different favourable condition for them to grow and sustain Insects have achieved enormous In terms of species diversity and abundance, it has been a success. Insects are the most abundant group of organisms on the planet, accounting for roughly 66% of all animal species. They can be found almost anywhere, and because they are excellent dispersers and exploiters of virtually all types of organic matter, they are an important part of every ecosystem and provide valuable ecosystem services. Insects have long been regarded as competitors in the fight for survival. Herbivorous insects cause 18% of global agricultural output to be lost. Despite this, pests make up less than 0.5% of the total number of bug species known. Humans controlling the environment choose crops for their larger growth and higher output. Jaiganesh is a type of insect pest. Sowmyashree and Srinivas [19] looked at the role of IOT in agriculture. It allows for the creation of yields, estimation, composts, diseases details for cure, and development proposals. Taneja et al. [20] is offering an irrigation system to help reduce water usage. They compute the amount of water by measuring various characteristics. The system is low-cost and energy-efficient. Sensors were employed to manage the irrigation valve, and a smart phone was used to monitor the situation remotely. Høye and Johanna [21] has introduced about the Computer vision and deep learning advances which may bring innovative answers to this global problem of accurate insect identification,. Entomological observations can be made effectively, continuously, and noninvasively using cameras and other sensors throughout the diurnal and seasonal cycles. In the lab, automated imaging can also capture the physical appearance of specimens. When trained on these data, many deep learning models can guess the nearby estimates of insect, biomass, and diversity. Deep learning models can also assess variation in phenotypic traits, behaviours, and interactions. To upgrade irrigation, Ratnayake et al. [22] Crop development, reduced water use, and proper water use are the goals. They calculated the required water quantity using humidity, temperature, and soil moisture. Ilyas et al. [23] Using computer vision and a deep learning network, the author attempts to observe insect behaviour in their natural habitat. They have developed a technique for tracking insect behaviour using imagebased tracking. a new hybrid detection and tracking method for outdoor monitoring of unmarked insects they developed a software and this software can detect an insect, discern when a tracked insect is blocked from view and when it re-emerges, and integrate a succession of insect locations into a coherent route. Nandyal et al. [24] has performed systematic literature review (SLR). To analyse and evaluate primary research of image-based insect identification and species classification algorithms, the author has deducted that 980 research published between 2010 and 2020 and chose 69 relevant studies using specified inclusion/exclusion criteria from among them. In this SLR, they examined the dataset properties (i.e. bug species targeted, crops, geographical locations, image capturing methods) and insect classification methodologies used in the primary studies.

A Practical Approach for Crop Insect Classification and Detection Using …

829

Jha et al. [25] plan a WSN for estimation of soil conditions. The kind of soil has evaluated and bases on soil composition the crop and fertilizers have prescribed. Thinking about the changing pace of soil, a methodology had applied to build a harmony between energy utilization and the exactness of phosphorus. Manoukis et al. [26] for entomologists, the author covers the principles of applied computer vision, including as image capture, data extraction, and analysis. They go over some of the most cutting-edge imaging gear and cameras, as well as lighting, software, and basic data collection scenarios, as well as specific examples. Quantification of behavioural events will become more widespread in insect research, that computer vision techniques for quantification will become more widely used, and that the application of these tools and approaches will yield new insights and answers to entomology challenges’. Alzu’b et al. [27] Using a case study, focuses on CloudIoT for solutions in various sectors. The MQTT protocol was used to create an irrigation system that is 22% more energy efficient and 15% faster. Tripathi et al. [28] aims at classifying soils based on characteristics and recommending the best crop using IoT and ML. Rehman et al. [29], used IoT to build a smart greenhouse. They propose voice control, which replaces the conventional interface, lowers the barrier of entry to science and technology, inspires a range of applications, and improves the production. Venkat et al. [30] proposes a geo-fencing and livestock tracking solution. Create a safe zone using IoT and GPRS, with dedicated sensors for the livestock. Cattle can be monitored and controlled remotely using data on their location, well-being, and health. Nagaveni et al. [31], a system has developed to watch the growth of crops development parameters. The system consists of a drone, with a camera to record images. It includes various crops decisions based on image. From images, they analysis the amount of green in leaf, moisture content etc. Sanches et al. [32] The Internet of Things (IoT), Wireless Communications (WiFi), Machine Learning (ML), and Artificial Intelligence (AI) are all explored. Crop diseases, storage management, insecticides, weed management, irrigation challenges, and water management are all problems in agriculture. Soil production and fertility have both been found to improve with automation. According to, Mas et al. [33] To meet problems, the agribusiness is armed itself with tools like sustainable farming. IoT is helpful in all aspects of farming, crop monitoring, water level monitoring, pest and animal control, and soil richness data. It has been determined to employ remote sensing methods. The situation was evaluated using remote cameras. Bjerge and co. The author demonstrates a portable computer vision system that can draw in and find live insects. The results suggest that live animals should be photographed and utilised to identify and categorise species when they are attracted to a light trap. An Automated Moth Trap (AMT) was constructed to draw and keep track of live insects during the evening and night. It has a camera and different light sources. Counting and Classifying Moths (MCC) a computer vision system based on deep learning image analysis, tracked and counted insects while also recognising moth species. 48 evenings, more than 250,000 people participated. An average of 5675 pictures were taken each night [34]. A special convolutional neural network was trained on 2000 labelled photos. Figueiredo et al. [35] The author of the study has suggested a smart trap with Internet of Things (IoT) capabilities

830

R. Yadav and A. Seth

that employs computer vision to recognise the desired insect. The answer consists of three parts: A web application that displays data using a programmable heat map, an embedded system with a camera, GPS sensor, and motor actuators, and IoT middleware that serves as a database service provider. The primary concern is addressed and the suggested cure is put into practise. Saoud [36] provide a novel method to assess the ease of identification of insects using their characteristic values (CV). They investigated two things in order to do this: (1) changes in SPIPOLL IESs, and (2) the connection between CV and IESs. The CV be applied to determine the IES of SPIPOLL insects, according to the findings.

3 Literature Summary The farmers’ world need to explore and test innovative technology to meet our food needs and create sustainable farming practises because of the expanding population, global warming, and quick changes in climatic conditions. On the other hand, farming is difficult and has a low income due to low production yields and uncontrollable variables. As a result, improving the processes involved in traditional farming is essential to improve crop output and quality. In this regard, IoT and machine learning-based methods significantly affect farming activities. Crop output and the management of agricultural resources are just two examples of how agriculture operations have improved. Numerous recent additions to agricultural activities have been examined, including monitoring, resource planning, irrigation (water supply), fertiliser, soil productivity, classification, disease identification, and others. These technologies enable farmers to obtain useful data for loss prevention while automating a variety of tasks. Application trends have also been found based on the literature. Figure 1 shows a variety of IoT and machine learning-based applications [37]. This increases the power of farmers and farming practises. According to current trends, irrigation and water management automation takes up the majority of work, with crop recommendation models and irrigation coming in second. Monitoring and examination of the soil are also important. Databased methodologies are beneficial [38]. The apps that also put a focus on animal welfare include the ones that develop management information systems (MIS), reviews, and surveys. Due to certain meteorological conditions, a particular form of pest is produced in the environment, which leads to a particular type of disease in the crop, lowering the quality of the crop thus result in degradation of the farmers as well as the country.

4 Objectives The focus of the work is to survey various methods of pest detection and classification in the field of agriculture using machine-learning algorithms along with use of IoT and propose a method to be included in our future work for pest detection and

A Practical Approach for Crop Insect Classification and Detection Using …

831

4%

Fig. 1 Applications of ML and IoT in agriculture [39]

MIS

12% 16%

Monitoring Soil

24%

16% 8%

20%

RecommendaƟons Animal IrrigaƟon Review

classification based on shape and size features using machine learning and IoT. In this context, based on known facts some key objectives has established as.

4.1 To Review and Understand the Morphology of the Insects To study the particular insect verities on the basis of kind of defect made by the insect whether the insect is a leaf eater or stem attacker or the insect attacks on the root also the different classes of the insects which grows in a particular whether system with the help of available literature and the data obtained from the field directly also to identify the specific disease associate with a specific kind of insect.

4.2 To Detect and Classify the Pest Based on Shape and Size Features, Using ML Techniques The aim is detect and classify pest using shape and size features data model that usage the data in the image format to train and predict the class of pest and soil dryness to provide instructions to cultivate the crops or other relevant treatment like fertilizer and pest control. This will help to enhance the involved agricultural processes.

4.3 To Study Various Machine Learning and Deep Learning Algorithm for Insect Classification and Identification To study about the various literatures available on the machine-learning algorithm for the detection and classification of insect.

832

R. Yadav and A. Seth

4.4 To Develop the Hybrid Machine-Learning Algorithm for the Classification and Detection of Insects In order to get the optimized results we will develop the hybrid version of the machinelearning algorithm.

5 Recent Study Recently, investigated some effective ML method of pest detection and classification. Thus, proposed a new machine-learning model for pest detection and classification in the crop fields with insect database, which are publically available [40, 41]. The established objectives are working on ML based techniques. In this context, an experimental study has carried out for identifying the efficient and accurate ML algorithms. In this study, algorithms namely, C4.5 decision tree, Support Vector Regression (SVR), Multi-layer Perceptron (MLP), Linear Regression (LR) and kNearest Neighbour (KNN) classifier was use. We have taken a dataset of different sizes from UCI machine learning repository. Additionally, we have considered two parameters i.e. accuracy and time to compare the models. Based on a set of experiments the obtained results are demonstrating in Fig. 2. The performance of the model in terms of accuracy has given in Fig. 2a. Figure 2b demonstrates the training time in MS. According to experimental observation, we found that the MLP, SVR, and KNN are producing higher accurate prediction [39]. On the other hand, in terms of time consumption the SVR, C4.5 decision tree, and linear regression is providing low time consumption. However, based on the performance of both the parameters we found MLP produces higher accuracy and less time consumption. Thus, we have proposed to be using the suitable variants of artificial neural network (ANN). Along with this we have studied the CNN model [42]. The CNN model is a type of neural network that lets us extract higher-level representations from picture input. Unlike classical image recognition, which requires the user to create image characteristics, CNN takes the raw pixel data from the image, trains the model, and then extracts the features for better categorization. And YOLO [43] Many applications, such as selfdriving cars, necessitate great precision and real-time inference speed. As a result, selecting an Object Detector that satisfies both speed and accuracy requirements are critical. YOLO (You Only Look Once) is a single-stage object detector that accomplishes both of these objectives (i.e., speed and accuracy). To offer you a complete view of the YOLO family, there are various YOLO versions (for example, YOLOv1, YOLOv2, YOLOX, YOLOR).

A Practical Approach for Crop Insect Classification and Detection Using …

833

Fig. 2 The comparison of machine learning algorithms in terms of a accuracy (%) and b training time [39]

6 Proposed Work Based on the conducted study and observed facts, we have established some objectives. The process work flow is shown in Fig. 3. Among them, we have discussed the objective to design a ML framework for support the agriculture process by suggesting the treatment for the crops by classification of different type of pest. The Hardware (Graphical processing Unit) for experiment to be carried out is Nvidia RTX 4000 with 1 TB of SSD.

6.1 Dataset and Methods For the future perspective we have performed our partial work in this paper for that at the very first stage we have collected the samples of insect images from IP102 dataset [44]. Dataset Exhibits natural Long tailed Distribution, It has hierarchical Taxonomy and the insect pests.

834

R. Yadav and A. Seth

Insect Image Data Set

Image Preprocessing

CalculaƟon of accuracy parameters

Feature Extracted

ApplicaƟ of Google Net & Alex Net

Fig. 3 The work flow

Which mainly affect one specific agricultural product is grouped in the same upper level category. Our proposed work consists of following data summary. The percentage of Train image is 79% and the Test image is about 21%. Train Image Count: 15,178 Test Image Count: 3798 Total Image Count: 18,976 ∼ =19,000. X-axis: Class Labels 0–102 Y-axis: total number of samples the Y-axis values are not shown clearly as the amount of data is very large. The sample images related to the datasets are shown in Fig. 4.

6.2 Image Preprocessing In order to get the exact features of the images we have applied following steps given below. All the images were randomly applied preprocessing, the preprocessing functions were applied in the following sequential way. 1. 2. 3. 4. 5. 6. 7.

Grayscale image with 3 channel output. Canny Edge Detection method applied on images. Cropping the image to the bounding boxes. Random Horizontal Flip with probability = 0.5. Random Rotation with degree of rotation = [−20, 20]. Random vertical Flip with probability = 0.3. Converting to tensor to be fed into Dataset Classes.

A Practical Approach for Crop Insect Classification and Detection Using …

835

Sample Images

Fig. 4 Sample images

The Noise reduction helped us in the correct identification of the edges of the insects images. For the Noise reduction, The above 7 steps has been applied on image data. X-axis: Class Labels. Y-axis: Total No of Samples. The training and the testing datasets are shown in Figs. 5 and 6.

836

R. Yadav and A. Seth

Fig. 5 Training data set

Fig. 6 Testing dataset

6.3 Features Extracted and Method of Extraction Dataset used is unstructured, and thus the number of features extracted is a black box in nature. Although the features extracted are directly correlated to the number of learnable params in each of the algorithms used. Initially Cross Validation was not applied while splitting the dataset; Dataset was split from the start. Till now what we observed is In Vision Based learning, moreover it increases variance in CNN networks after applying Cross Validation.

6.4 Methodology The methodology framework are shown in Figs. 7 and 8. Rescaled Size of image: H ∗W ∗D

A Practical Approach for Crop Insect Classification and Detection Using …

837

Fig. 7 Alex Net [45]

Fig. 8 GoogleNet [46]

H Height: 227 W Width: 227 D Image channels: 3 (all are same as the image is grayed out after transformations).

6.5 Filter Size for Each Layer Filter size are shown in Figs. 9 and 10.

6.6 Classification Accuracy Classification accuracy for Alex net and GoogleNet are shown in Table 1.

Fig. 9 Filter size for Alex Net [45]

838 R. Yadav and A. Seth

Fig. 10 Filter size for Google Net [46]

A Practical Approach for Crop Insect Classification and Detection Using …

839

840

R. Yadav and A. Seth

Table 1 Classification accuracy for Alexnet and GoogleNet Model

Best val loss

Best accuracy

Precision

Recall

F1 score

Alexnet

48.1019

0.3431

1

1

0.5

8.5205

0.3532

1

1

0.5

GoogleNet

Table 2 Correct identification of insect class Predicted class

Class name

Actual class

Class name

4

Asiatic rice bowler

4

Asiatic rice bowler

Table 3 Wrong identification of insect class Predicted class

Class name

Actual class

Class name

52

Blister beetle

51

Legume blister beetle

6.7 Computational Time Both the classifiers were trained with zero-shot learning and thus training time was relatively fast. Alexnet: ~15 s for 10 epochs GoogleNet: ~16 s for 10 epochs Highest Classification Accuracy Alex net = 0.3431 GoogleNet = 0.3532 (Table 2). GoogleNet showed better classification Accuracy (Table 3). Figures 11, 12 and 13 are the validated output results. Classifiers are trained to give an array of probabilities with 102 elements signifying the distinct 102 classes of the dataset. The achieved Confidence AlexNet and Google Net are 98.34 and 99.34 respectively. The Loss value is 48.10 and 8.520 respectively.

7 Conclusion In this paper Study on number of research paper has been done about the insect classification and detection using the deep learning model and the existing dataset which has been collected by the no of different resources like IoT, Google, manual etc. This paper shows an experiment which lead us towards the bigger solution for the problem of insect identification and classification specially for the Indian farmers. Further work can be done using several other deep learning models like Yolo6 from where further results can be deduced for insect identification and classification.

A Practical Approach for Crop Insect Classification and Detection Using …

Fig. 11 Training time graph

Fig. 12 Confidence 0.98345

Fig. 13 Confidence 0.99342

841

842

R. Yadav and A. Seth

References 1. Abeywardhana DL, Dangalle CD, Nugaliyadde A et al (2022) An ultra-specific image dataset for automated insect identification. Multimedia Tools Appl 81:3223–3251 2. Nanni L, Manfè A, Maguolo G, Lumini A, Brahnam A (2022) High performing ensemble of convolutional neural networks for insect pest image detection. Ecol Inf 67:101515. ISSN: 1574-9541 3. Bhattacharya P, Neamtiu I, Shelton CR (2012) Automated, highly-accurate, bug assignment using machine learning and tossing graphs. J Syst Softw 85(10):2275–2292. ISSN: 0164-1212 4. Rajasekaran T, Anandamurugan S (2019) Challenges and applications of wireless sensor networks in smart farming—a survey. Adv Bi Da Cl Comp Adv Inte Sys Comp 750 5. Maduranga MWP, Abeysekera R (2020) Machine learning applications in IOT based agriculture and smart farming: a review. Int J Engg App Sci Tech 4(12):24–27 6. Sowmiya M, Prabavathi S (2019) Smart agriculture using Iot and cloud computing. Int J Rec Tech Engg 7(6S3) 7. Abd El-GhanyShadia NM, Abd El-AzizShadia E, Abd El-AzizShahira E, Marei S (2020) Environ Sci Pollut Res 27(6) 8. Utkin KY, Marenych M, Galych O, Sliusar I (2020) Main aspects of the creation of managing information system at the implementation of precise farming. In: 11th IEEE international conference on depe. system, service and technology, Kyiv, Ukraine 9. Gimeno CR, Voort Mvd, Niemid JK, Lauwers L, Kristensen AR, Wauters E (2019) Assessment of the value of information of precision livestock farming: a conceptual framework. NJAS –Wage. J Li. Sci 90–91:100311 10. Farooq MS, Riaz S, Abid A, Abid K, Naeem MA (2019) A survey on the role of IoT in agriculture for the implementation of smart farming, vol 7. IEEE, spec. sect. on new technology for Sm. Far. 4.0: research challenge & opportunity 11. Doshi J, Patel T, Bharti Sk (2019) Smart farming using IoT, a solution for optimally monitoring farming conditions. Proc Comput Sci 160:746–751 12. Rekha P, Ramesh MV, Rangan VP, Nibi KV (2017) High yield groundnut agronomy: an IoT based precision farming framework. 978-1-5090-6046-7/17/$31.00 ©2017. IEEE 13. Jaiganesh S, Gunaseelan K, Ellappan V (2017) IOT agriculture to improve food and farming technology. In: Proceedings of IEEE conference on emerging devices and smart systems, Mahe. Engg. Coll., Tamilnadu, India. 978-1-5090-5555-5/17/$31.00. IEEE 14. Wu X, Zhan C, Lai Y-K, Cheng M-M, Yang J (2019) Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), pp 8787–8796 15. Thenmozhi K, Srinivasulu Reddy U (2019) Crop pest classification based on deep convolutional neural network and transfer learning. Comput Electron Agric 164:104906. ISSN 0168-1699 16. Caballero D, Calvini R, Manuel Amigo J (2020) Hyperspectral imaging in crop fields: precision agriculture. Amigo JM (ed) Data handling in science and technology, vol 32. Elsevier 17. Kulkarni A, Kulkarni P, Dandin P, Kumar S (2020) Smart and economic farming using IoT. © JUL 2020, IRE J 4(1) 18. Jankielsohn A (2018) The importance of insects in agricultural ecosystem. Adv Entomol 6(2) 19. Sowmyashree S, Srinivas C (2020) A survey on smart soil analysis and predicting the irrigation using IOT a literature survey and review paper. Int J Inn Sci Res Tech 5(2) 20. Taneja M, Jalodia N, Byabazaire J, Davy A, Olariu C (2019) SmartHerd management: a microservices-based fog computing–assisted IoT platform towards data-driven smart dairy farming. Softw: Pract Exper 49:1055–1078. published by John Wiley & Sons Ltd. 21. Høye T, Johanna A (2022) Deep learning and computer vision will transform entomology. In: Proceedings of the national academy of sciences. https://doi.org/10.1073/pnas.2002545117 2022/03/27 22. Ratnayake MN, Dyer AG, Dorin, Tracking individual honeybees among wildflower clusters with computer vision-facilitated pollinator monitoring. PLoS ONE 16(2):e0239504. https:// doi.org/10.1371/journal.pone.0239504

A Practical Approach for Crop Insect Classification and Detection Using …

843

23. Ilyas QM, Ahmad M (2020) Smart farming.: an enhanced pursuit of sustainable remote livestock tracking and geofencing using IoT and GPRS. Hind Wire Comm Mob Comp 2020, Art. ID 6660733:12 24. Nandyal S, Khamitkar PS, Joshi PS (2019) Agridrone: automation of agriculture using IoT. Int J Innov Sci Res Tech 4(6) 25. Jha K, Doshi A, Patel P, Shah M (2019) A comprehensive review on automation in agriculture using artificial intelligence. Artif Intel Agric 2:1–12 26. Manoukis NC, Collier TC (2019) Computer vision to enhance behavioral research on insects. Ann Entomol Soc Am 112(3):227–235. https://doi.org/10.1093/aesa/say062 27. AlZu’bi S, Hawashin B, Mujahed M, Jararweh Y, Gupta BB, An efficient employment of internet of multimedia things in smart and future agriculture. Mult Appl. https://doi.org/10. 1007/s11042-019-7367-0 28. Tripathi MK, Maktedar DD (2020) A role of computer vision in fruits and vegetables among various horticulture products of agriculture fields: a survey. Inf Proc Agric 7:183–203 29. Rehman TU, Mahmud MS, Chang YK, Jin J, Shin J (2019) Current and future applications of statistical machine learning algorithms for agricultural machine vision systems. Comput Electr Agric 156:585–605 30. Venkat PS, Avinash BL, Jabber B (2020) Crop yield prediction based on Indian agriculture using machine learning. In: International conference for emerging technology, Belgaum, India 31. Nagaveni V, Raghavendra BK (2019) A review on machine learning classification techniques for plant disease detection. In: 5th International conference on advanced computer and communication system, 978-1-5386-9533-3/19/$31.00 ©2019. IEEE 32. Sanches D, Júnior AL, da Costa CC, de Castro Victoria D, Inamasu RY, Grego CR, Ferreira VR, Ramirez AR (2020) Precision and digital agriculture: adoption of technologies and perception of Brazilian farmers. Agriculture 10:653. https://doi.org/10.3390/agriculture10120653 33. Más FR (2020) From smart farming towards agriculture 5.0.: a review on crop data management. Agronomy 10:207. https://doi.org/10.3390/agronomy10020207. www.mdpi.com/journal/agr onomy 34. Bjerge KN, Sepstrup JB, Helsing-Nielsen MV, Høye F (2021) An automated light trap to monitor moths (Lepidoptera) using computer vision-based tracking and deeplearning. Sensors. https://doi.org/10.3390/s21020343 35. Figueiredo V, Campos A, Mafra S, Rodrigues J (2020) A proposed IoT smart trap using computer vision for sustainable pest control in coffee culture. arXiv 2020 36. Saoud Z (2020) Can we estimate insect identification ease degrees from their identification key paths. Ecol Inf 55:101010. ISSN: 1574-9541 37. Geetha K (2021) An integrated approach for crop production analysis from geographic information system data using SqueezeNet. J Soft Comput Paradigm 4:308–321 38. Krishnan GH, Rajasenbagam T (2021) A comprehensive survey for weed classification and detection in agriculture lands. J Inf Technol 3(4):281–289 39. Yadav R, Seth A (2022) A review for investigation on soil features using IoT and ML. In: 2022 4th international conference on smart systems and inventive technology (ICSSIT) 40. Abeywardhana DL, Dangalle CD, Nugaliyadde A et al (2022) An ultra-specific image dataset for automated insect identification. Multimed Tools Appl 81:3223–3251.https://doi.org/10. 1007/s11042-021-11693-3 41. Espinoza K, Valera DL, Torres JA, López A, Molina-Aiz FD (2016) Combination networks as a novel approach for the identification of Bemisia tabaci and Frankliniella of image processing and artificial neural occidentalis on sticky traps in greenhouse agriculture. Comput Electron Agric 127:495–505 42. https://towardsdatascience.com/understanding-cnn-convolutional-neural-network-69fd62 6ee7d4 43. Sharma A (2022) Introduction to the YOLO family. In: Chakraborty D, Chugh P, Gosthipaty AR, Haase J, Huot S, Kidriavsteva K, Raha R, Thanki A (eds) PyImageSearch. https://pyimg. co/3cpmz

844

R. Yadav and A. Seth

44. Wu X, Zhan C, Lai Y-K, Cheng M-M, Yang J (2019) IP102.: a large-scale benchmark dataset for insect pest recognition. In: 2019 IEEE/CVF conference on computer vision and pattern recognition, pp 8779–8788. https://doi.org/10.1109/CVPR.2019.00899 45. Iandola FN, Han S, Moskewicz MW, Ashraf K, Dally WJ, Keutzer K, SqueezeNet.: AlexNetlevel accuracy with 50x fewer parameters and